- Split View
-
Views
-
Cite
Cite
Omid Haji-Ghassemi, Ryan J Blackler, N Martin Young, Stephen V Evans, Antibody recognition of carbohydrate epitopes, Glycobiology, Volume 25, Issue 9, September 2015, Pages 920–952, https://doi.org/10.1093/glycob/cwv037
- Share Icon Share
Abstract
Carbohydrate antigens are valuable as components of vaccines for bacterial infectious agents and human immunodeficiency virus (HIV), and for generating immunotherapeutics against cancer. The crystal structures of anti-carbohydrate antibodies in complex with antigen reveal the key features of antigen recognition and provide information that can guide the design of vaccines, particularly synthetic ones. This review summarizes structural features of anti-carbohydrate antibodies to over 20 antigens, based on six categories of glyco-antigen: (i) the glycan shield of HIV glycoproteins; (ii) tumor epitopes; (iii) glycolipids and blood group A antigen; (iv) internal epitopes of bacterial lipopolysaccharides; (v) terminal epitopes on polysaccharides and oligosaccharides, including a group of antibodies to Kdo-containing Chlamydia epitopes; and (vi) linear homopolysaccharides.
Introduction
Many of the concepts that define modern molecular immunology stem from studies of antibody recognition of carbohydrate antigens, chiefly carried out by pioneers such as Drs Michael Heidelberger and Elvin Kabat on pneumococcal polysaccharides, dextrans and blood group antigens. However, the structural characterization of carbohydrate-specific antibodies was then hampered by the lack of homogeneous functional antibody species. Nevertheless, work on heterogeneous preparations by Kabat and others established many key characteristics, e.g., that the combining site could accommodate up to six residues and could take the form of a pocket or groove (Kabat 1978). The general structure of antibodies (Figure 1) was established by protein sequencing and crystallography in the 1970s, and fully elucidated by the structures of two whole IgGs determined by McPherson and colleagues (Harris et al. 1998). The discovery that mouse myelomas can secrete homogeneous IgA and IgM anti-carbohydrate antibodies (Potter and Leon 1968) led to a major advance in understanding antibody–antigen recognition with the solution of two crystal structures of Fab fragments: the phosphocholine-binding IgA M603 (Satow et al. 1986) and the galactan-binding IgA J539 (Suh et al. 1986). The M603 Fab structure included phosphocholine bound in a small pocket in the centre of the binding site and since phosphocholine can be a dominant epitope when it occurs on polysaccharides (Young et al. 2013), M603 may be considered the first anti-carbohydrate-antibody crystal structure.
The specificities of myeloma immunoglobulins were determined empirically, but the introduction of hybridoma technology by Kohler and Milstein (1975) made available monoclonal antibodies to a wide range of defined carbohydrate antigens. The resulting developments in understanding the antibodies produced in response to carbohydrate antigens have been comprehensively reviewed by Brorson et al. (2002). On the structural side, the first hybridoma Fabs solved with bound antigen fragments were Se155-4 (Cygler et al. 1991), an antibody to a Salmonella lipopolysaccharide (LPS), and BR96 (Jeffrey et al. 1995), an antibody to a human tumor antigen, the Lewis Y antigen (Ley). Their binding sites showed the grooves and cavities that had been predicted by Kabat (Wilson and Stanfield 1995). These antibodies represent the two major types of carbohydrate antigen that have subsequently been structurally investigated.
Immunology of carbohydrate antigens
The numerous possible combinatorial linkages, modifications and relative degree of flexibility of many carbohydrates require antibodies to utilize a variety of strategies in their recognition. Details of the genetic events that lead to the observed binding site diversity of antibodies are described in several reviews (Dudley et al. 2005; Maizels 2005; Jung et al. 2006).
In summary, the majority of the observed immunoglobulin array stems from the formation of a nascent B-cell lymphocyte, where genes coding for one variable heavy (Vh) and one variable light (VL) domain are constructed from a limited repertoire of inherited germ-line gene segments. These consist of V (variable), D (diversity) and J (joining) gene segments for the heavy chain, and VJ segments for the light chain that are located on a separate chromosome and can be either kappa or lambda type. The recombination events can be quite variable, and substitutions are often incorporated between the gene segments that result in productive and non-productive immunoglobulins. V(D)J recombined genes encoding VH and VL domains are further paired with constant (C) gene segments that determine antibody isotype. Following translation, the antibody polypeptides are modified at glycosylation sites, particularly in the constant-regions (Figure 1) where these carbohydrate moieties are involved in modulation of effector functions (Mimura et al. 2001; Jefferis 2009). The initial result in nascent B-cells is a class M immunoglobulin glycoprotein of specific sequence, and each B-lymphocyte will display multiple copies of its particular membrane-bound IgM antibody on its surface.
The final genetic source of antibody diversity relies on T-cell help. If the antigen is T-cell-dependent like many proteins and peptides, T-cell co-stimulation of the B-cell can induce somatic hypermutation of the antibody genes to produce daughter cells with mutant immunoglobulins of potentially higher affinity (Sharon 1990; Li et al. 2004; Teng and Papavasiliou 2007).
Generally speaking, carbohydrates are classified as T-cell independent antigens (Murphy et al. 2012), where B-lymphocytes are activated without the presentation of antigen fragments via MHC molecules to T-lymphocytes (Vos et al. 2000). However, the inability of most carbohydrate antigens to recruit T-cell help results in a B-cell response lacking affinity maturation and weighted toward the production of IgM and IgG2 in human and IgM and IgG3 in mouse (Greenspan et al. 1988; Scott et al. 1988; Stein 1992; Ullrich 2009; Wigelsworth et al. 2009). Furthermore, the anti-carbohydrate immune response usually produces antibodies with “V-region restriction” where a relatively limited set of germ-line gene segments will generate antibodies against a broad range of epitopes (Pascual et al. 1992; Brorson et al. 2002; Nguyen et al. 2003; Brooks, Müller-Loennies, et al. 2010; Blackler et al. 2012).
To overcome this restricted response, glycoconjugate antigens have been developed in which carbohydrate antigens or fragments thereof are coupled to proteins, and the protein moieties can then recruit T-cell help. The intracellular processing of carbohydrate antigens has been found to involve degradation by means of reactive oxygen and nitrogen species (Duan and Kasper 2011), and this knowledge has led to improved designs for a second generation of glycoconjugate vaccines (Buskas et al. 2008; Astronomo and Burton 2010; Avci et al. 2011; Costantino et al. 2011). However, two exceptions to the T-independent paradigm have been found: Polysaccharides that carry both negatively and positively charged substituents, i.e., zwitterions, can interact with MHCII species (Avci and Kasper 2010). The oxidative breakdown of polysaccharide antigens can also produce species capable of this type of interaction (Velez et al. 2009). Secondly, some glycolipid antigens are presented by MHC homologs CD1a, b, c and d, to various families of T-cell receptors (TCRs; Icart et al. 2008).
Affinities of anti-carbohydrate antibodies
The observed affinities of anti-carbohydrate antibodies are typically lower by factors of 103–105 than antibodies specific for protein or peptide antigens (Krause and Coligan 1979; MacKenzie et al. 1996; Brorson et al. 2002). This is compensated by their initial expression as decavalent IgM and their observed class switching bias toward IgG3 in mice and IgG2 in humans, which tend to self-associate through their constant regions to form multivalent networks (Greenspan et al. 1988; Cooper et al. 1991). The multivalent nature of these antibody clusters results in a marked increase in avidity (Edberg et al. 1972; Greenspan and Cooper 1992; Yoo et al. 2003) and reflects an evolved mechanism for the recognition of multivalent or densely displayed carbohydrate antigens. The surface clustering of multivalent antibodies can only occur when there are correspondingly large numbers of antigen molecules present, which serves to distinguish between cells that display many copies of the antigen (such as bacteria) and normal cells that display just a few. This ability of antibodies to distinguish chemically identical epitopes depending on their environment is termed “context dependent recognition” (Ramos and Moller 1978; Wylie et al. 1982; Caoili 2010), and is particularly relevant for tumor antigens.
The lower affinities observed for carbohydrate-specific antibodies, and other carbohydrate-binding proteins such as lectins, derive from the binding not being driven only by enthalpic factors, and emphasise the relative importance of entropic considerations (Bundle and Young 1992; Bundle et al. 1998; Engström et al. 2005). The general lack of rigidly defined structures in many carbohydrates would require entropically unfavourable immobilization of these otherwise flexible segments upon antibody binding. However, attempts to demonstrate this experimentally have had mixed results; a rigid antigen analog of a Salmonella epitope tethered between the O-6 atoms of the Gal and Man residues was bound equally well as the free form (Bundle et al. 1998), while a similar analog of a Shigella flexneri epitope showed enhanced binding affinity (McGavin and Bundle 2005). Additionally, the extension of a ligand by the addition of a sugar unit can lead to ambiguity in the interpretation of binding data, as the fixing of the anomeric oxygen atom in one conformation may increase affinity without the added sugar necessarily contacting the antibody surface. In contrast, if the new sugar unit has cooperative interactions with the protein and enthalpic contributions that offset the loss of conformational entropy, the overall Ka will be similar to that of the shorter substrate (Kabat 1960; Jennings 2012). Such misinterpretation can be minimized by studying each oligosaccharide in its methyl glycoside (MAG) form, which also avoids α/β mutarotation equilibrium that may complicate ligand modelling in structural studies (Bundle 1989; Yates et al. 1996; Bundle et al. 2012).
The hydrophilic nature of carbohydrates increases the possibility that water molecules have to be displaced or trapped during complex formation, both of which have distinct entropic consequences, making the net thermodynamic contribution of each interaction difficult to model. Generally, a greater desolvation of receptor and ligand corresponds to higher affinity, as the inherent entropic penalty of carbohydrate binding is offset (Woods 1998; Fadda and Woods 2010). While the role of the water molecules that solvate both the carbohydrate ligand and its receptor was considered in detail in the 1990s (Chervenak and Toone 1994; Lemieux 1996), this area is still not fully understood, particularly the role of hydrophobic interactions (Snyder et al. 2011). The overall low binding of carbohydrates is usually accompanied by relatively fast kon and koff rates, summarized for lectins in Scharenberg et al. (2014). Kinetic data have been reported for only a few of the anti-carbohydrate antibodies described in this review. Mutants of the anti-blood group A single-chain fragment variable (scFv), blood group A (BGA; Thomas et al. 2002) had kon rates of 2.4–5.4 × 104 M−1 s−1 and koff rates of 3.7–6.4 × 10−2 s−1 (KD values 0.8–2.1 µM) for the trisaccharide antigen, which are similar to those for the binding of Glc1Man9GlcNAc2 by calreticulin, 3.9 × 104 M−1 s−1 and 8 × 10−2 s−1 (Patil et al. 2000). Antibodies to charged carbohydrate antigens, such as the Kdo trisaccharide recognized by anti-Chlamydia antibodies, can have KD values <1 µM, with both kon and koff rates that are faster, e.g., 2 × 105 M−1 s−1 and 0.12 s−1 for S25-2 (Müller-Loennies et al. 2000).
Finally, the presence of –COOH or –CH3 groups on the carbohydrate can lead to higher affinities by permitting ionic or hydrophobic interactions, e.g., for antibodies that recognize charged Kdo [All carbohydrate nomenclature used are spelled out unless they are specified in the essentials of Glycobiology 2nd edition abbreviations list (Varki et al, 2009)] species (Müller-Loennies et al. 2000) and methyl groups on the Vibrio cholera Ogawa antigen (Villeneuve et al. 2000).
Antibody structure determination
To date, only a few structures of intact immunoglobulins have been determined (Harris et al. 1998; Saphire et al. 2003). The flexibility of the hinge region of an intact immunoglobulin (Harris et al. 1998) makes crystallization difficult. Structural studies are therefore generally carried out on Fab fragments, generated by limited proteolysis of the immunoglobulins with papain or pepsin. However, crystallizing Fabs with bound ligand is made difficult not only by the low affinities of anti-carbohydrate antibodies but also by the packing of Fab molecules in crystals where the constant region of one Fab often lies across the binding site of a neighbouring one, blocking the access of ligands. This mode of crystal contact is unfortunately only too common in Fab crystallizations (Davies et al. 1990). Nevertheless, nearly all the structures described here were obtained with Fabs from mouse hybridoma proteins, mostly IgG2a, or in the case of anti-HIV, from human hybridomas obtained from lymphocytes of resistant individuals. Structures of antibodies against carbohydrates deposited in the protein data bank can be accessed using SAbDab database (Dunbar et al. 2014) (http://opig.stats.ox.ac.uk/webapps/abdb/web_front/Welcome.php) or Glyco3D database (Pérez et al. 2015) (http://glyco3d.cermav.cnrs.fr/home.php).
Recombinant expression of smaller antigen-binding fragments, such as scFvs, in Escherichia coli can provide access to binding fragments from IgMs that cannot be obtained in useful quantities by proteolysis (Patenaude et al. 1998). It also enables concomitant studies by site-directed mutagenesis. Fragments consisting of only Vh or Vl domains, called single domain antibodies (sdAbs), are even smaller molecules that can be solved directly with NMR techniques (Vranken et al. 2002), though carbohydrate-specific sdAbs are rare (Stijlemans et al. 2004; El Khattabi et al. 2006; Behar et al. 2009).
Modelling of carbohydrate-antibody systems
Because many anti-carbohydrate antibodies could not be crystallized or could only be solved without bound ligand, modelling studies have often been necessary. In a series of landmark papers, Chothia and co-workers first demonstrated that the conformation of individual complementary determining loops (CDRs) can be grouped into just a few “canonical forms” based on the loop length and position/identity of key residues (Chothia and Lesk 1987; Chothia et al. 1989, 1992; Tramontano et al. 1990), and this work has since been expanded (Al-Lazikani et al. 1997; Morea et al. 1997, 1998; Abhinandan and Martin 2008; North et al. 2011). Structural prediction based on this information is limited by the difficulty in modelling the relative orientation of each CDR or even of the heavy and light chain domains (Reczko et al. 1995), and the highly variable nature of the CDR H3 (Kuroda et al. 2008). Further, contribution of solvent is generally ignored in such studies. On-line prediction packages are available that compare the query sequence to antibodies of known structure and select the closest homologs for each CDR (Marcatili et al. 2008; Sircar et al. 2009); the most effective application is reported to be Rosetta Antibody (Almagro et al. 2011).
The prediction of structures of antigen complexes is even more challenging, but the number of in silico binding and docking studies has been rapidly expanding with advances in theory and computing power (Woods 1998; Agostino et al. 2009, 2010; Fadda and Woods 2010; North et al. 2011). These methods have grown increasingly powerful, particularly when combined with NMR experiments. These are carried out in solution and can provide detailed and highly relevant information about the changes in the conformations of carbohydrate ligands as they bind (Peters and Pinto 1996; Haselhorst et al. 2009; Oberli et al. 2010). The most common experiments focus on 1H nuclear Overhauser effect transfer resonances and saturation transfer difference (STD)-NMR, which are sensitive to protein–carbohydrate interactions and to changes in protein and carbohydrate conformations upon binding (Kogelberg et al. 2003; Theillet et al. 2009; Roldós et al. 2011). Suitable on- and off-rates for the antigen–antibody interaction are critical for successful use of these methods, which otherwise may result in poor signal-to-noise ratios, thus making it difficult to distinguish the free from the bound state (Oberli et al. 2010). A striking example of what can be achieved by combining modelling, NMR and docking approaches is the complex structure prediction of an antibody with a β1,2Man antigen from Candida albicans (Johnson et al. 2012).
Types of carbohydrate antigen
It has been claimed that the only absolute rule in biology is that cell surfaces are festooned with carbohydrates (Gagneux and Varki 1999): glycoproteins and glycolipids in the case of eukaryotic cells, and capsular polysaccharides and LPS in the case of bacterial cells. These antigens are the most common classes studied in carbohydrate–antibody complexes. The eukaryotic carbohydrates include tumor antigens and glycoprotein structures that viruses acquire from their host cells’ biosynthetic machinery. The relative abundances of some mammalian surface carbohydrates have been observed to change when a cell becomes malignant, and these modified sugars, often on gangliosides, have been extensively investigated as potential targets for immunotherapy (Ragupathi 1996). These glycans are of course self-antigens and the immune responses to them can therefore lead to antibodies with remarkable structural properties.
The main antibodies that have been studied structurally (Table I) are directed towards bacterial carbohydrate epitopes found on the LPS and on mycobacterial lipoarabinomannans (LAMs). There is some crossover between eukaryotic and bacterial carbohydrate antigens because some bacteria evade immune surveillance by using antigenic mimicry, i.e., their polysaccharide antigens resemble carbohydrates normally found in human tissues. An example of this phenomenon occurs with the capsular polysaccharides of group B meningococci and E. coli K1 strain, which consist largely of homopolymers of α2,8-linked sialic acid (Frosch et al. 1985; Silver et al. 1988). The same structure occurs as a developmental antigen in fetal mammalian cells (Rosenberg et al. 1986; Rutishauser and Jessell 1988) and in neural cell adhesion molecule (Finne et al. 1983), and so is poorly immunogenic (Jennings and Lugowski 1981; Bartoloni et al. 1995).
Antibody . | Antigena . | Liganded PDB code . | Unliganded PDB code(s) . | Organism or antigen . | Antigen class . |
---|---|---|---|---|---|
2G12 | Man9GlcNAc2 | 1OP5 | 1OM3 | HIV-1 | HIV-1 glycoprotein |
PG9 | V1/V2 domain of gp120 | 3U4E | 3U36 | HIV-1 | HIV-1 glycoprotein |
PG16 | V1/V2 domain of gp120 | 4DQO | 3MUG | HIV-1 | HIV-1 glycoprotein |
PGT121 | Self N-glycan | 4FQC | 4FQ1, 4FQQ | HIV-1 | HIV-1 glycoprotein |
PGT124 | V3 domain of gp120 | 4R2G | 4R26 | HIV-1 | HIV-1 glycoprotein |
PGT128 | V3 domain of gp120 | 3TYG | HIV-1 | HIV-1 glycoprotein | |
PGT135 | Gp120 core | 4JM2 | 4JM4 | HIV-1 | HIV-1 glycoprotein |
chBR96 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1CLY | 1UCB | Ley antigen | Tumour epitopes |
hu3S193 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1S3K | Ley antigen | Tumour epitopes | |
291-2G3-A | Galβ1,4(Fucα1,3)GlcNAcβ (Lex) | 1UZ8 | 1UZ6 | Lex antigen | Tumour epitopes |
chR24 | Unliganded | 1BZ7 | GD3 glycosphingolipid | Tumour epitopes | |
P3 | Unliganded | 3IU4 | NeuGc-GM3 | Tumour epitopes | |
237mAb | ERGT(GalNAc)KPPPLEELS | 3IET | Tn antigen | Tumour epitopes | |
L363 | Mouse CD1d receptor and C20:2 αGal ceramide | 3UBX | Gal ceramide | Glycolipid | |
BGA | Unliganded | 1JV5 | Blood group A | Human ABO(H) blood group A | |
Se155-4 | Abeα1,3Manα1,4Rhaα1,3Gal | 1MFC | Salmonella B | Internal epitopes of bacterial LPS | |
SYA/J-6 | Rhaα1,2Rhaα1,3Rhaα1,3GlcNAcα1,2Rao | 1M7I | 1M71 | Sh. flexneri Y | Internal epitopes of bacterial LPS |
F22-4 | →2Rhaα1,2Rhaα1,3(Glcα1,4)Rhaα1,3GlcNAcβ1→ | 3BZ4 | 3C5S | Sh. flexneri 2a | Internal epitopes of bacterial LPS |
Ab52 | Unliganded | 3UJT | F. tularensis | Internal epitopes of bacterial LPS | |
N62 | Unliganded | 4KPH | F. tularensis | Internal epitopes of bacterial LPS | |
WN1 222-5 | GlcNα1,2Glcα1,2Glcα1,3(Galα1,6)Glcα1,3(Hepα1,7)(P4)Hepα1,3(P4)Hepα1,5(Kdoα2,4)Kdoα2,6(P4)GlcNβ1,6GlcNα(1P) | 3V0W | 3V0V | LPS core | Internal epitopes of bacterial LPS |
LPT3-1 | GlcNAcα1,2Hepα1,3(Glcβ1,4)Hepα1,5Kdo | 4C83 | N. meningitidis | Internal epitopes of bacterial LPS | |
S-20-4 | α1,2 2-O-Methylperosamine disaccharide (Ogawa) | 1F4Y | 1F4W | Terminal bacterial epitopes | |
S25-2 | Kdo2,8Kdo2,4Kdo | 3SY0 | 1Q9K, 1Q9L | Chlamydia | Terminal bacterial epitopes |
S25-2 | Kdo | 3T4Y | Chlamydia | Terminal bacterial epitopes | |
S54-10 | Kdo2,4Kdo2,4Kdo | 3I02 | Chlamydia | Terminal bacterial epitopes | |
S73-2 | Kdo2,8Kdo2,4Kdo | 3HZV | Chlamydia | Terminal bacterial epitopes | |
S67-27 | Kdo2,8-7-O-Me-Kdo | 3IKC | Chlamydia | Terminal bacterial epitopes | |
S64-4 | Kdo2,8Kdo2,4Kdo2,6GlcN4P1,6GlcN1P | 3PHO | Chlamydia | Terminal bacterial epitopes | |
S25-26 | Kdo2,8Kdo2,4Kdo-O-allyl | 4M7J | 4M7Z, 4M93, 4MA1 | Chlamydia | Terminal bacterial epitopes |
CS-35 | Araβ1,2Araα1,5(Araβ1,2Araα1,3)Araα1,5Araα | 3HNS | Mycobacteria | Terminal bacterial epitopes | |
J539 | Unliganded | 2FBJ | Galactan | Homopolysaccharide epitopes | |
Yst9.1 | Unliganded | 1MAM | B. abortus | Homopolysaccharide epitopes | |
mAb735 | α2,8NeuAc (sialic acid) octamer | 3WBD | Polysialic acid | Homopolysaccharide epitopes |
Antibody . | Antigena . | Liganded PDB code . | Unliganded PDB code(s) . | Organism or antigen . | Antigen class . |
---|---|---|---|---|---|
2G12 | Man9GlcNAc2 | 1OP5 | 1OM3 | HIV-1 | HIV-1 glycoprotein |
PG9 | V1/V2 domain of gp120 | 3U4E | 3U36 | HIV-1 | HIV-1 glycoprotein |
PG16 | V1/V2 domain of gp120 | 4DQO | 3MUG | HIV-1 | HIV-1 glycoprotein |
PGT121 | Self N-glycan | 4FQC | 4FQ1, 4FQQ | HIV-1 | HIV-1 glycoprotein |
PGT124 | V3 domain of gp120 | 4R2G | 4R26 | HIV-1 | HIV-1 glycoprotein |
PGT128 | V3 domain of gp120 | 3TYG | HIV-1 | HIV-1 glycoprotein | |
PGT135 | Gp120 core | 4JM2 | 4JM4 | HIV-1 | HIV-1 glycoprotein |
chBR96 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1CLY | 1UCB | Ley antigen | Tumour epitopes |
hu3S193 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1S3K | Ley antigen | Tumour epitopes | |
291-2G3-A | Galβ1,4(Fucα1,3)GlcNAcβ (Lex) | 1UZ8 | 1UZ6 | Lex antigen | Tumour epitopes |
chR24 | Unliganded | 1BZ7 | GD3 glycosphingolipid | Tumour epitopes | |
P3 | Unliganded | 3IU4 | NeuGc-GM3 | Tumour epitopes | |
237mAb | ERGT(GalNAc)KPPPLEELS | 3IET | Tn antigen | Tumour epitopes | |
L363 | Mouse CD1d receptor and C20:2 αGal ceramide | 3UBX | Gal ceramide | Glycolipid | |
BGA | Unliganded | 1JV5 | Blood group A | Human ABO(H) blood group A | |
Se155-4 | Abeα1,3Manα1,4Rhaα1,3Gal | 1MFC | Salmonella B | Internal epitopes of bacterial LPS | |
SYA/J-6 | Rhaα1,2Rhaα1,3Rhaα1,3GlcNAcα1,2Rao | 1M7I | 1M71 | Sh. flexneri Y | Internal epitopes of bacterial LPS |
F22-4 | →2Rhaα1,2Rhaα1,3(Glcα1,4)Rhaα1,3GlcNAcβ1→ | 3BZ4 | 3C5S | Sh. flexneri 2a | Internal epitopes of bacterial LPS |
Ab52 | Unliganded | 3UJT | F. tularensis | Internal epitopes of bacterial LPS | |
N62 | Unliganded | 4KPH | F. tularensis | Internal epitopes of bacterial LPS | |
WN1 222-5 | GlcNα1,2Glcα1,2Glcα1,3(Galα1,6)Glcα1,3(Hepα1,7)(P4)Hepα1,3(P4)Hepα1,5(Kdoα2,4)Kdoα2,6(P4)GlcNβ1,6GlcNα(1P) | 3V0W | 3V0V | LPS core | Internal epitopes of bacterial LPS |
LPT3-1 | GlcNAcα1,2Hepα1,3(Glcβ1,4)Hepα1,5Kdo | 4C83 | N. meningitidis | Internal epitopes of bacterial LPS | |
S-20-4 | α1,2 2-O-Methylperosamine disaccharide (Ogawa) | 1F4Y | 1F4W | Terminal bacterial epitopes | |
S25-2 | Kdo2,8Kdo2,4Kdo | 3SY0 | 1Q9K, 1Q9L | Chlamydia | Terminal bacterial epitopes |
S25-2 | Kdo | 3T4Y | Chlamydia | Terminal bacterial epitopes | |
S54-10 | Kdo2,4Kdo2,4Kdo | 3I02 | Chlamydia | Terminal bacterial epitopes | |
S73-2 | Kdo2,8Kdo2,4Kdo | 3HZV | Chlamydia | Terminal bacterial epitopes | |
S67-27 | Kdo2,8-7-O-Me-Kdo | 3IKC | Chlamydia | Terminal bacterial epitopes | |
S64-4 | Kdo2,8Kdo2,4Kdo2,6GlcN4P1,6GlcN1P | 3PHO | Chlamydia | Terminal bacterial epitopes | |
S25-26 | Kdo2,8Kdo2,4Kdo-O-allyl | 4M7J | 4M7Z, 4M93, 4MA1 | Chlamydia | Terminal bacterial epitopes |
CS-35 | Araβ1,2Araα1,5(Araβ1,2Araα1,3)Araα1,5Araα | 3HNS | Mycobacteria | Terminal bacterial epitopes | |
J539 | Unliganded | 2FBJ | Galactan | Homopolysaccharide epitopes | |
Yst9.1 | Unliganded | 1MAM | B. abortus | Homopolysaccharide epitopes | |
mAb735 | α2,8NeuAc (sialic acid) octamer | 3WBD | Polysialic acid | Homopolysaccharide epitopes |
aDifferent nomenclature is used for antigens consisting of protein/peptide and carbohydrates.
Antibody . | Antigena . | Liganded PDB code . | Unliganded PDB code(s) . | Organism or antigen . | Antigen class . |
---|---|---|---|---|---|
2G12 | Man9GlcNAc2 | 1OP5 | 1OM3 | HIV-1 | HIV-1 glycoprotein |
PG9 | V1/V2 domain of gp120 | 3U4E | 3U36 | HIV-1 | HIV-1 glycoprotein |
PG16 | V1/V2 domain of gp120 | 4DQO | 3MUG | HIV-1 | HIV-1 glycoprotein |
PGT121 | Self N-glycan | 4FQC | 4FQ1, 4FQQ | HIV-1 | HIV-1 glycoprotein |
PGT124 | V3 domain of gp120 | 4R2G | 4R26 | HIV-1 | HIV-1 glycoprotein |
PGT128 | V3 domain of gp120 | 3TYG | HIV-1 | HIV-1 glycoprotein | |
PGT135 | Gp120 core | 4JM2 | 4JM4 | HIV-1 | HIV-1 glycoprotein |
chBR96 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1CLY | 1UCB | Ley antigen | Tumour epitopes |
hu3S193 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1S3K | Ley antigen | Tumour epitopes | |
291-2G3-A | Galβ1,4(Fucα1,3)GlcNAcβ (Lex) | 1UZ8 | 1UZ6 | Lex antigen | Tumour epitopes |
chR24 | Unliganded | 1BZ7 | GD3 glycosphingolipid | Tumour epitopes | |
P3 | Unliganded | 3IU4 | NeuGc-GM3 | Tumour epitopes | |
237mAb | ERGT(GalNAc)KPPPLEELS | 3IET | Tn antigen | Tumour epitopes | |
L363 | Mouse CD1d receptor and C20:2 αGal ceramide | 3UBX | Gal ceramide | Glycolipid | |
BGA | Unliganded | 1JV5 | Blood group A | Human ABO(H) blood group A | |
Se155-4 | Abeα1,3Manα1,4Rhaα1,3Gal | 1MFC | Salmonella B | Internal epitopes of bacterial LPS | |
SYA/J-6 | Rhaα1,2Rhaα1,3Rhaα1,3GlcNAcα1,2Rao | 1M7I | 1M71 | Sh. flexneri Y | Internal epitopes of bacterial LPS |
F22-4 | →2Rhaα1,2Rhaα1,3(Glcα1,4)Rhaα1,3GlcNAcβ1→ | 3BZ4 | 3C5S | Sh. flexneri 2a | Internal epitopes of bacterial LPS |
Ab52 | Unliganded | 3UJT | F. tularensis | Internal epitopes of bacterial LPS | |
N62 | Unliganded | 4KPH | F. tularensis | Internal epitopes of bacterial LPS | |
WN1 222-5 | GlcNα1,2Glcα1,2Glcα1,3(Galα1,6)Glcα1,3(Hepα1,7)(P4)Hepα1,3(P4)Hepα1,5(Kdoα2,4)Kdoα2,6(P4)GlcNβ1,6GlcNα(1P) | 3V0W | 3V0V | LPS core | Internal epitopes of bacterial LPS |
LPT3-1 | GlcNAcα1,2Hepα1,3(Glcβ1,4)Hepα1,5Kdo | 4C83 | N. meningitidis | Internal epitopes of bacterial LPS | |
S-20-4 | α1,2 2-O-Methylperosamine disaccharide (Ogawa) | 1F4Y | 1F4W | Terminal bacterial epitopes | |
S25-2 | Kdo2,8Kdo2,4Kdo | 3SY0 | 1Q9K, 1Q9L | Chlamydia | Terminal bacterial epitopes |
S25-2 | Kdo | 3T4Y | Chlamydia | Terminal bacterial epitopes | |
S54-10 | Kdo2,4Kdo2,4Kdo | 3I02 | Chlamydia | Terminal bacterial epitopes | |
S73-2 | Kdo2,8Kdo2,4Kdo | 3HZV | Chlamydia | Terminal bacterial epitopes | |
S67-27 | Kdo2,8-7-O-Me-Kdo | 3IKC | Chlamydia | Terminal bacterial epitopes | |
S64-4 | Kdo2,8Kdo2,4Kdo2,6GlcN4P1,6GlcN1P | 3PHO | Chlamydia | Terminal bacterial epitopes | |
S25-26 | Kdo2,8Kdo2,4Kdo-O-allyl | 4M7J | 4M7Z, 4M93, 4MA1 | Chlamydia | Terminal bacterial epitopes |
CS-35 | Araβ1,2Araα1,5(Araβ1,2Araα1,3)Araα1,5Araα | 3HNS | Mycobacteria | Terminal bacterial epitopes | |
J539 | Unliganded | 2FBJ | Galactan | Homopolysaccharide epitopes | |
Yst9.1 | Unliganded | 1MAM | B. abortus | Homopolysaccharide epitopes | |
mAb735 | α2,8NeuAc (sialic acid) octamer | 3WBD | Polysialic acid | Homopolysaccharide epitopes |
Antibody . | Antigena . | Liganded PDB code . | Unliganded PDB code(s) . | Organism or antigen . | Antigen class . |
---|---|---|---|---|---|
2G12 | Man9GlcNAc2 | 1OP5 | 1OM3 | HIV-1 | HIV-1 glycoprotein |
PG9 | V1/V2 domain of gp120 | 3U4E | 3U36 | HIV-1 | HIV-1 glycoprotein |
PG16 | V1/V2 domain of gp120 | 4DQO | 3MUG | HIV-1 | HIV-1 glycoprotein |
PGT121 | Self N-glycan | 4FQC | 4FQ1, 4FQQ | HIV-1 | HIV-1 glycoprotein |
PGT124 | V3 domain of gp120 | 4R2G | 4R26 | HIV-1 | HIV-1 glycoprotein |
PGT128 | V3 domain of gp120 | 3TYG | HIV-1 | HIV-1 glycoprotein | |
PGT135 | Gp120 core | 4JM2 | 4JM4 | HIV-1 | HIV-1 glycoprotein |
chBR96 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1CLY | 1UCB | Ley antigen | Tumour epitopes |
hu3S193 | Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal (Ley) | 1S3K | Ley antigen | Tumour epitopes | |
291-2G3-A | Galβ1,4(Fucα1,3)GlcNAcβ (Lex) | 1UZ8 | 1UZ6 | Lex antigen | Tumour epitopes |
chR24 | Unliganded | 1BZ7 | GD3 glycosphingolipid | Tumour epitopes | |
P3 | Unliganded | 3IU4 | NeuGc-GM3 | Tumour epitopes | |
237mAb | ERGT(GalNAc)KPPPLEELS | 3IET | Tn antigen | Tumour epitopes | |
L363 | Mouse CD1d receptor and C20:2 αGal ceramide | 3UBX | Gal ceramide | Glycolipid | |
BGA | Unliganded | 1JV5 | Blood group A | Human ABO(H) blood group A | |
Se155-4 | Abeα1,3Manα1,4Rhaα1,3Gal | 1MFC | Salmonella B | Internal epitopes of bacterial LPS | |
SYA/J-6 | Rhaα1,2Rhaα1,3Rhaα1,3GlcNAcα1,2Rao | 1M7I | 1M71 | Sh. flexneri Y | Internal epitopes of bacterial LPS |
F22-4 | →2Rhaα1,2Rhaα1,3(Glcα1,4)Rhaα1,3GlcNAcβ1→ | 3BZ4 | 3C5S | Sh. flexneri 2a | Internal epitopes of bacterial LPS |
Ab52 | Unliganded | 3UJT | F. tularensis | Internal epitopes of bacterial LPS | |
N62 | Unliganded | 4KPH | F. tularensis | Internal epitopes of bacterial LPS | |
WN1 222-5 | GlcNα1,2Glcα1,2Glcα1,3(Galα1,6)Glcα1,3(Hepα1,7)(P4)Hepα1,3(P4)Hepα1,5(Kdoα2,4)Kdoα2,6(P4)GlcNβ1,6GlcNα(1P) | 3V0W | 3V0V | LPS core | Internal epitopes of bacterial LPS |
LPT3-1 | GlcNAcα1,2Hepα1,3(Glcβ1,4)Hepα1,5Kdo | 4C83 | N. meningitidis | Internal epitopes of bacterial LPS | |
S-20-4 | α1,2 2-O-Methylperosamine disaccharide (Ogawa) | 1F4Y | 1F4W | Terminal bacterial epitopes | |
S25-2 | Kdo2,8Kdo2,4Kdo | 3SY0 | 1Q9K, 1Q9L | Chlamydia | Terminal bacterial epitopes |
S25-2 | Kdo | 3T4Y | Chlamydia | Terminal bacterial epitopes | |
S54-10 | Kdo2,4Kdo2,4Kdo | 3I02 | Chlamydia | Terminal bacterial epitopes | |
S73-2 | Kdo2,8Kdo2,4Kdo | 3HZV | Chlamydia | Terminal bacterial epitopes | |
S67-27 | Kdo2,8-7-O-Me-Kdo | 3IKC | Chlamydia | Terminal bacterial epitopes | |
S64-4 | Kdo2,8Kdo2,4Kdo2,6GlcN4P1,6GlcN1P | 3PHO | Chlamydia | Terminal bacterial epitopes | |
S25-26 | Kdo2,8Kdo2,4Kdo-O-allyl | 4M7J | 4M7Z, 4M93, 4MA1 | Chlamydia | Terminal bacterial epitopes |
CS-35 | Araβ1,2Araα1,5(Araβ1,2Araα1,3)Araα1,5Araα | 3HNS | Mycobacteria | Terminal bacterial epitopes | |
J539 | Unliganded | 2FBJ | Galactan | Homopolysaccharide epitopes | |
Yst9.1 | Unliganded | 1MAM | B. abortus | Homopolysaccharide epitopes | |
mAb735 | α2,8NeuAc (sialic acid) octamer | 3WBD | Polysialic acid | Homopolysaccharide epitopes |
aDifferent nomenclature is used for antigens consisting of protein/peptide and carbohydrates.
Antibodies to glycan epitopes on HIV
A major goal of HIV vaccine development has been directed towards induction of broadly neutralizing monoclonal antibodies that recognize epitopes in the envelope proteins (McGaughey et al. 2004; Pantophlet and Burton 2006; Pinter 2007; Haynes et al. 2010), with increasing focus on the glycoconjugate epitopes of HIV-1 (Wang 2006). The variable region of gp120 has long been a challenge for structural studies due to its sequence heterogeneity and its abundant glycosylation with high-mannose structures (Man9GlcNAc2). Referred to as the glycan shield, it displays self-antigens of the host and acts as a physical barrier to protect the HIV envelope trimer against immunoglobulin recognition (Cao et al. 1997; Rusert et al. 2011). The antibodies described in these studies were obtained from three HIV-1-infected patients. Notably, they have structures that differ significantly from the usual form of immunoglobulin shown in Figure 1, including domain-swapped structures and unusually long CDR loops. Structural work on antibodies to glycan epitopes on gp120 began with characterization of the binding of high-mannose oligosaccharides, then progressed through designed fragments of gp120 bearing two N-glycans (which can be mannose-rich or hybrid types), then complete gp120, and finally to a trimer of the envelope protein with antibody PGT 122. The latter showed both its interactions with the glycans and the relationship of the gp41 and gp120 subunits (Julien et al. 2013). The antibodies described here use three different types of structure to access distinct glycopeptide epitopes on gp120, which overlap around a “hotspot” at Asn332 (Kong et al. 2013).
The domain-swapped antibody, 2G12
The human monoclonal antibody 2G12 broadly neutralizes T-cell line-adapted strains of HIV-1 through activation of both antibody-dependent cell-mediated cytotoxicity and complement system (Buchacher et al. 1994; Trkola et al. 1996), with a strong dependence on the multiplicity of N-linked glycans on the surface of the glycoprotein gp120 (Trkola et al. 1996).
The structure of 2G12 Fab complexed with Man9GlcNAc2 (Figure 2A) reveals arguably the most remarkable example of high avidity antibody binding to carbohydrate antigens (Calarese et al. 2003, 2005), where the Vh domains of two neighbouring Fabs “swap” to form a Fab dimer (Kunert et al. 1998). The swapping of the Vh domains produces an additional antigen-combining site at the Vh/Vh interface that allows 2G12 to bind a cluster of glycans present on the surface of gp120 separated by as much as 35 Å. The critical nature of the domain swap was demonstrated using mutagenesis studies of recombinant full-length 2G12 antibody, which showed that a single mutation Ile19 (Figure 2A) to Arg on the Vh domain resulted in a non-domain-swapped variant of 2G12 that was able to bind Manα1,2Man motifs presented on an ELISA plate, but unable to recognize Manα1,2Man clusters in cells expressing the HIV-1 envelope trimer (Doores et al. 2010). The Manα1,2Man ends of the three arms of the Man9 structure are the main sites recognized by 2G12 (Figure 2B), particularly the D1 and D2 arms (Calarese et al. 2005).
2G12 displays an intriguing similarity to the binding mechanism of the cyanobacterial lectin, cyanovirin, which is also capable of neutralizing HIV-1 through binding to the D1 arm of Man9GlcNAc2 on surface of gp120 (O'Keefe et al. 2000) and has a domain-swapped form similar to that of 2G12 (Botos et al. 2002).
Anti-HIV antibodies with long CDR H3
There are a number of conventional (non-domain swapped) broadly neutralizing antibodies in addition to 2G12 that display specificity for mixed N-linked glycans on gp120 of HIV-1 (Pejchal et al. 2011). Monoclonal antibodies PG9 and PG16 both bind the V1/V2 region to neutralize a broad range of HIV-1 strains with high efficacy, however, the variable nature of gp120 further complicates structural analysis.
To overcome the structural heterogeneity of gp120 and aid crystallization trials, McLellan and colleagues generated glycopeptide antigen mimics that retained the minimal scaffold of the V1/V2 region. Complexed structures of V1/V2 scaffolds from two different HIV-1 strains with PG9 were subsequently determined to 2.19 and 1.80 Å resolution. The structures revealed another unusual binding mechanism in which a very long CDR H3 of 26 residues with a charged “hammerhead” tip inserts into the glycan shield, fitting neatly between two N-linked glycans to contact the protein surface of gp120 (Figure 2C; Sattentau 2011). In this way, PG9 overcomes sequence variability inherent to the V1/V2 protein by recognition of the protein backbone rather than to the variable amino acid side chains (Figure 2D). The CDR H3 contacts not only two N-glycans but also forms parallel β-sheet like bonds to a β-strand of the antigen. The tip of the loop is highly unusual in that it includes sulfated Tyr residues, which can form salt-bridges to cationic residues on the antigen. Because the CDR H3 penetrates so deeply, it contacts the inner Asn-linked GlcNAc as well as terminal Man residues of the Man5GlcNAc2 (Asn160) (Figure 2D). Interactions with a second N-glycan at Asn156 or Asn173 are also formed by Tyr100K of CDR H3 and the inner GlcNAc. The strong evolutionary conservation of the two glycans in gp120 allows PG9 to recognize some 80% of HIV-1 strains (McLellan et al. 2011).
A second antibody PG16 was solved in complex with the V1/V2 scaffold to 2.44 Å resolution, which showed many shared binding characteristics of PG9 (Figure 2E), including binding the Asn160 glycan, the H-bonding to the peptide backbone and the charged interactions of the sulfated Tyr on a 26-residue CDR H3 (Pancera et al. 2013). However, the second N-glycan at Asn173 was of the complex type, with one Neu5Ac-Gal-GlcNAc arm on the Man5GlcNAc2 core (Figure 2F). Interactions with the α2,6-linked Neu5Ac dominated PG16's binding, contributing half the surface area and most of the H-bonds to the antibody. The residues that form the contacts are different in PG9, explaining its preference for Man5GlcNAc2. Three PG16-specific amino acids: ArgL94, SerL95 and HisL95A form the complementary surface that allow for the recognition of the terminal sialic acid (Figure 2F), whereas in PG9 the three amino acids are ThrL94, ArgL95 and ArgL95A. Replacement of the three CDR L3 residues of PG9 with that of PG16 resulted in a chimeric antibody that displayed specificity for both mannose rich and complex-type N-glycans; leading to enhanced HIV-1 neutralization (Pancera et al. 2013).
Another example of an antibody with a long CDR H3 of 22 residues, PGT 121 was solved in a “self-complex” in which one Fab was bound to the complex biantennary glycan of a second Fab (Figure 2G; Mouquet et al. 2012). This crystal form was obtained during an attempted co-crystallization with a non-sialylated biantennary glycan. The sialic acid was again α2,6-linked and formed the bulk of the contacts with the antibody, which were exclusively from the Vh domain (Figure 2H). Recently, the structure of a closely related antibody PGT 124 was solved in complex with gp120 (Garces et al. 2014). PGT 124 facilitate binding to HIV trimer via a single glycosylation site (Figure 2I), while all other gp120-specific antibodies discussed here utilize multiple glycans for high-affinity interaction. The authors reason that this mechanism has evolved from distinct affinity maturation events; whereby PGT 124 is selected for efficient recognition of the HIV trimer by minimizing binding to surrounding heterogeneous N-glycans (Garces et al. 2014). Despite the smaller recognized surface of this antibody, its neutralizing potential is comparable with other anti-HIV antibodies (Sok et al. 2014).
Anti-HIV antibodies with additional CDR insertions
Monoclonal antibodies PGT 127 and PGT 128 share the preference shown by 2G12 for the Manα1,2Man ends of the D1 and D2 arms of high-mannose glycans (Figure 2J), but are structurally distinct (Pejchal et al. 2011). As well as 17-residue long CDR H3s, they have six residue insertions in their H2 CDRs that are critical for glycan recognition. Structures with Man9 were obtained at higher resolutions (1.65 and 1.29 Å) than for any other anti-gp120 antibodies. Hence the H-bonding network and the involvement of water molecules were well defined. The recognition was through a surface defined by CDRs H2, H3 and L3; key residues included four Trp (Figure 2K), two being from the CDR H2. The buried surface areas were high, 748 Å2 for PGT 128, consistent with the higher viral neutralization titers of these antibodies compared with PG9 (648 Å2) (Kong et al. 2014). A structure was also obtained with an engineered gp120 outer domain bearing two N-linked glycans, for which the Ka was 2.2 × 107 M−1. Both glycans and part of the peptide were bound, with a Man9GlcNAc2 structure in the previous Man9 site and a Man5GlcNAc2 portion of an N-glycan in a second site. The latter is unusual in being formed not only by the longer CDR H2 but also by residues 72–75 of the framework region 2 or sometimes referred to as the CDR H4 loop. At the same time, CDR H3 contacts the peptide backbone of the antigen.
Another unusual CDR has been found in PGT 135, which has a five-residue insertion in CDR H1 along with an 18 residue long CDR H3 (Kong et al. 2013). Its structure was solved in unliganded form and with an assemblage of other molecules, namely gp120 core, CD4 and an anti-CD4 Fab (Kong et al. 2013). No large conformational changes were seen between the two forms. The two long CDRs protrude from the antibody surface far enough to match the length of the glycans, and reach side chains of the protein (Figure 2L). The total size of the peptide-glycan epitope is very large, 1334 Å2, with 70% contributed by the glycans (Kong et al. 2014). Endoglycosidase treatment of a PGT 135-gp120 complex showed protection of N-glycans at three sites, though the Fab contacts two of the glycans at Asn392 and Asn332 (Figure 2M) more than the third at Asn386. The Asn392 and Asn332 glycans interact with opposite faces of CDR H3 in a bifurcated manner, from the ends of the D2 and D3 arms down to the GlcNAc residues. PGT 135 recognizes the opposite face of the Asn332 glycan compared with PGT 128 and consequently there are fewer H-bonds and more apolar interactions (Figure 2M), including ones with a sequence of seven hydrophobic residues of CDR H3.
Antibodies to tumor carbohydrate epitopes
Some of the most intriguing and potentially clinically important studies of anti-carbohydrate antibodies stem from the observation that some cell-surface oligosaccharides are tumor antigens, such as the Lewis x and y antigens, their sialylated counterparts, and gangliosides. Gangliosides are glycosphingolipids, which are ceramide-linked oligosaccharides with at least one terminal sialic acid residue. The modulation of ganglioside type and concentration has been associated with the growth and differentiation of tissues, and with carcinogenesis (Hakomori and Kannagi 1983; Hakomori 1989, 2001; Bitton et al. 2002; Birkle et al. 2003; Xu et al. 2005). Hence, cancer vaccines based on glycolipids are receiving increasing attention (Wandall and Tarp 2007; Durrant et al. 2012) including some completely synthetic ones (Buskas et al. 2008; Wilson and Danishefsky 2013).
One of the best-studied tumor-related antigens is a single GalNAc-α residue O-linked to Ser or Thr, referred to as the Tn antigen (Ju et al. 2011; Richichi et al. 2014). The Tn antigen is present in many different types of cancer and generally not present in normal adult tissues (Ju et al. 2011). Consequently, it offers a perfect target for immunotherapy and has been the subject of many antibody studies (Takahashi et al. 1988; Numata et al. 1990; Baldus et al. 1992; Hakomori 2001; Brooks, Schietinger et al. 2010; Yuasa et al. 2012).
The Ley-specific antibodies, BR96 and 3S193
The Ley surface antigen is expressed at high levels in a variety of tumor cell lines, and as a result antibodies against this structure have been intensively investigated in hopes of coupling the fragment variable (Fv) and Fab fragments with toxic substances or radiochemicals as “magic bullets” that selectively target and kill malignant cells (Hellström and Hellström 1991; Oldham and Dillman 2008). Its structure is Fucα1,2Galβ1,4(Fucα1,3)GlcNAcβ1,3Gal-R.
The monoclonal antibody BR96 specific to the Ley antigen showed promising results during phase I clinical trials; the humanized Fab conjugated with the anti-cancer drug doxorubicin was rapidly internalized by cancer cells, resulting in tumor clearance in nude mice (Hellström et al. 1990; Garrigues et al. 1994). While phase II trials were underway, the crystal structures of both the murine and humanized Fab structures were solved in complex with a nonoate methyl ester glycoside derivative of Ley, to 2.80 and 2.50 Å, respectively (Chang et al. 1994; Jeffrey et al. 1995). The BR96 antibody binds the Ley tetrasaccharide from its non-reducing end (Figure 3A) in a large cavity 10 Å deep and 12 Å wide. This is formed mainly by aromatic residues (Figure 3B) and binding energy is postulated to stem from hydrophobic effects. The buried surface area was 422 Å2, most of it from the Vh domain whose CDR H3 also formed most of the H-bonds. Though specific contacts are made to all four sugars, the affinity of the Fab is only 2 × 105 M−1 (Yelton et al. 1995). A comparison with the unliganded form of the antibody, determined at 2.6 Å (Sheriff et al. 1996), showed evidence for an induced fit since CDRs L1, L2 and H2 changed conformations on binding. However, CDR H3 which makes the major contacts with the antigen did not change conformation significantly.
A randomized phase II study of BR96-doxorubicin conjugate completed in 1999 resulted in significant gastrointestinal toxicities in patients with metastatic breast cancer, probably due to the presence of Ley antigen at lower levels in the gastrointestinal tract of normal epithelial cells. As a result of these studies, BR96–doxorubicin conjugate was not approved for human use.
A second Ley-specific antibody 3S193 has been generated using mice immunized with human adenocarcinoma cells (Kitamura et al. 1994). It displayed no cytotoxicity to O, A, AB and B human blood group containing erythrocytes (Kitamura et al. 1994; Scott et al. 2000), making it a suitable therapeutic candidate. A 1.9 Å resolution crystal structure of the humanized 3S193 has been determined (Ramsland et al. 2004). Nucleotide sequence comparison between BR96 and 3S193 revealed that they share V-genes and not surprisingly, the 3S193 antibody formed a similarly shaped pocket (Figure 3C) dominated by hydrophobic residues (Figure 3D). Despite these similarities, there were nine amino acid differences compared with BR96 in the CDR regions, seven of which were located on CDRs H2 and H3. Further, the orientation of the main chain backbone of CDR L1 allows humanized 3S193 to form a hydrogen bond to the Ley-specific Fuc residue through Asn28 (Figure 3D), whereas this residue is further away in BR96. In contrast to the BR96 structure, there are numerous water molecules in the combing site of 3S193 that participate in bridging interactions.
The Lex-specific antibody 291-2G3-A
The Lex antigen is not only expressed on tumor cells but also occurs on the surface of parasitic worms such as schistosomes. The structure of 291-2G3-A Fab was solved with and without the trisaccharide Galβ1,4(Fucα1,3)GlcNAcβ (Van Roon et al. 2004), with a Ka measured for this hapten of 9.3 × 104 M−1. The liganded structure was obtained to 2.05 Å by soaking the unliganded crystals with the Lex trisaccharide, which was possible due to solvent channels in the crystal allowing access to the combining site (Van Roon et al. 2004). There is a shallow binding pocket 15× 13 × 10 Å (Figure 3E), in which the trisaccharide can contact all six of the CDRs, with a buried surface of 302 Å2. As with many carbohydrate-specific antibodies, several hydrophobic side chains are oriented towards the hydrophobic face of the pyranose ring and the methyl group of N-acetylglucosamine MAG, providing the driving force needed for binding (Figure 3F). Unlike many antibodies, however, 291-2G3-A involves all six CDRs in binding, forming highly specific interactions with Lex antigen. Comparison between bound and unbound states indicated a lock and key binding mechanism.
The ganglioside GD3-specific antibody, R24
Murine IgG3 antibody R24 is specific for the ganglioside GD3, Neuα2,8Neuα2,3Galβ1,4Glcβ-ceramide, and displays one of the most sophisticated mechanisms for carbohydrate recognition by an antibody. GD3 is a normal differentiation antigen and is observed in low concentrations on cell surfaces; however, GD3 can be found in high concentrations on the surfaces of melanomas, soft tissue sarcomas and tumors (Dippold et al. 1980; Hamilton et al. 1993). R24 has been shown to recognize membrane surfaces containing high concentrations of GD3 while ignoring lower concentrations, making R24 a potential venue for cancer therapy (Urmacher et al. 1989; Helfand et al. 1999).
The crystal structure of unliganded R24 was determined at 3.1 Å resolution for the mouse Fab and at 2.5 Å for a chimeric form with human constant domains (Kaminski et al. 1999). A prominent pocket, 8.5 × 12 × 8 Å, is formed by the H1, H2 and H3 loops, which is large enough to accommodate the terminal sialic acid residue of GD3 (Figure 3G). R24 binds GD3 very weakly in solution, yet it has been observed to bind to cell-surface GD3 with higher affinities than can be explained by bivalent IgG binding. Part of this binding can be explained by the natural tendency of murine IgG3 antibodies to associate through their respective constant regions to form multivalent antibody clusters (Greenspan et al. 1988; Berney et al. 1991; Greenspan and Cooper 1992). In addition, R24 has been shown to bind other molecules of R24 through their antigen-binding domains via “homophilic binding”, which occurs while R24 is binding membrane-anchored GD3 (Kaminski et al. 1999) (Figure 3H). This means that R24 contains an idiotope not only for GD3 (IDGD3) but also an idiotope to other molecules of R24 through homophilic binding (IDHOM). The large number of cooperative interactions forming on the surface increases the likelihood of complement activation and recruitment of other immune effector cells for the eventual lysis of the cancerous cell.
Site-directed mutagenesis studies have shown that IDGD3 and IDHOM are distinct sites, as mutants could be generated which removed IDHOM without disrupting antigen binding. The location of these amino acid residue mutations suggested that CDR H2 was the location of IDHOM (Chapman et al. 1990). This was confirmed in the crystal structure of the unliganded Fab from R24, which showed that homophilic binding proceeded through a β-strand interaction between the H2 loops of adjacent R24 molecules (Kaminski et al. 1999). Further, R24 self-association through its murine IgG3 effector regions results in the formation of multivalent arrays on the cell surface displaying high concentrations of GD3, which explains why R24 ignores cell surfaces with GD3 oligosaccharides present in normal levels. R24 provides one of the best examples of antibody recognition of antigen in a context-specific manner.
The ganglioside NeuGc-GM3-specific antibody, P3
The P3 mAb was generated by immunizing mice with the ganglioside N-glycolyl-GM3 (NeuGcα2,3Galβ1,4Glcβ-ceramide), which is often present on the surface of malignant cells (Vazquez 1995). P3 binds a range of N-glycosyl-containing gangliosides as well as sulfatides, and an anti-idiotype mAb of P3, named 1E10, has been used successfully to produce antibodies against these tumor-associated antigens in clinical trials (Vazquez et al. 1998; Alfonso et al. 2002; Guthmann et al. 2006; Neninger et al. 2007; Hernandez et al. 2008). To prevent cross-reactivity with the acetylated variant of the antigen (NeuAc, lacking only an oxygen atom compared with the NeuGc), a chimeric P3 (chP3) was produced, which is remarkably specific to the NeuGc-GM3/GM2 gangliosides.
Modelling and mutagenesis analysis of the CDR regions of P3 revealed surface complementary in shape and charge to a NeuGc residue, with polar contacts to positively charged arginine residues located in this groove (Perez et al. 2001; Lopez-Requena, De Acosta, et al. 2007). A single mutation of an Arg residue in CDR H1 abolished binding to the ganglioside antigen (Lopez-Requena, Rodriguez, et al. 2007).
As with R24 antibody, the liganded structure of chP3 is not available; however, the unliganded Fab structure of chP3 was solved to 1.75 Å resolution (Talavera et al. 2009). It has a long CDR H3 loop that protrudes from the middle of the site, separating the Vl and Vh domains, and hence there is no central cavity. Unfortunately, CDR H3 was not accurately modeled in the crystal structure posing further complications for binding prediction. In order to use docking stimulations one needs to overcome the diversity of conformations available for CDR H3 and the flexibility of the ligand. Talavera's group used nine different conformations for the CDR H3 backbone and a set of criteria to predict the complex structure. These criteria included the mutational and binding assay data for chP3 and the energy and conformational restrictions of the ligand. As a result, they were able to narrow down 1800 possible different liganded structures to only one conformation, which fit their binding and mutational data. The model displays a hydrophilic pocket with all functional groups of NeuGc participating in interactions with the antibody. The hydrophilic nature of the binding site would also likely involve water bridged interactions, which are harder to predict during docking simulations and were therefore excluded from this study. Moreover, the model complemented earlier binding data with derivatives of NeuGc-GM3, which anticipated that the carboxylate group of NeuGc would be facing the guanidinium group of Arg31 Vh chain, forming a salt bridge with this residue (Moreno et al. 1998).
The Tn antigen-specific antibody, 237mAb
The Tn antigen and related T-antigen are among the most studied cancer carbohydrate antigens, and the Fab from 237mAb, an IgG2a κ, was the first glycan-specific anti-Tn antibody to be solved. Its structure was determined with a dodecameric GalNAc-peptide, ERGT(GalNAc)KPPPLEELS, a segment of a tumor-associated glycoprotein, at 2.2 Å resolution, and with free GalNAc at 2.6 Å (Brooks, Schietinger, et al. 2010). The antibody is not observed to bind the unglycosylated peptide, and the GalNAc residue dominates the interaction with glycopeptide (Figure 3I and J). The Ka for the glycopeptide is 7.1 × 106 M−1, a high value for a carbohydrate ligand but not unusual for a peptide. The site takes the form of a shallow groove for the peptide moiety, with the GalNAc accommodated in a central pocket formed by germ-line residues of CDRs H2, H3 and L1. The buried surface area was 532 Å2. Every hydroxyl group forms at least one H-bond directly to the antibody, without involvement of water bridges. Overall, there are 8 H-bonds from the Vl domain and 11 from the Vh. Thus specificity for the tumor antigen is achieved by a combination of interactions with both the glycan and peptide moieties.
Other mammalian glycan epitopes
The αGal ceramide-specific antibody, L363
Antibodies that mimic TCR binding are useful to study or modulate T-cell function by specifically blocking TCR activation (Mareeva et al. 2008; Dahan et al. 2011; Dahan and Reiter 2012). When the glycolipid α-galactosyl ceramide (αGal ceramide) is presented to T-helper cells though CD1d receptor of invariant natural killer T-cells, the latter become activated and secrete cytokines for increased anti-tumor activity (Van Kaer 2004; Hong and Park 2007). This property of αGal ceramide has spurred investigators to study the interaction of the TCR with the CD1d molecule using antibodies against the αGal ceramide in its presented form (Yu et al. 2005). One antibody called L363 was found to bind αGal ceramide glycolipid only when bound to CD1d, resembling biding to the TCR (Yu et al. 2012). The binding affinity (dissociation constant, KD) was measured to be in the nanomolar range by surface plasmon resonance (SPR).
The structure of L363 Fab bound to mouse CD1d-αGal Ceramide glycolipid analogue determined to 3.1 Å resolution (Figure 3K) reveals contacts to the antigen similar to the TCR, though the individual CDRs do not superimpose well with the loops of the TCR (1–3α and 1–3β) (Yu et al. 2012). Binding to the CD1d-glycolipid complex requires both L and H chains, where contacts to glycolipid occur only with the L-chain (Figure 3L) and the H-chain forms a surface complementary to the CD1d, forming similar contacts as the TCR. A crucial contact occurs between the mouse CD1d receptor and Leu99 of the TCR 3α loop, which is replaced by the Trp104 residue in CDR H3 of L363 antibody. The structure highlights the significant recognition of an αGal residue in conjunction with key residues of CD1d receptor in order to achieve mimicry of the TCR, though in contrast to the TCR, L363 antibody is not able to induce the structural changes required to bind glycolipids from Borrelia and Streptococcus species.
The blood group A-specific antibody, BGA
It is well known that the immune response to blood group trisaccharide antigens is strong, and a mismatched blood transfusion can lead to massive intravascular hemolysis and eventual death. However, the A and B blood group antigens differ only in the substitution of a hydroxyl group for an N-acetyl group on the terminal sugar residue. The A and B trisaccharide antigens are, respectively: GalNAcα1,3(Fucα1,2)Galβ-O-R and Galα1,3(Fucα1,2)Galβ-O-R, where R is a carbohydrate moiety of a glycolipid or glycoprotein. The recognition of these epitopes is therefore extremely specific. The ABO blood groups have been shown to contain “micro-epitopes”, where different antibodies recognize unique orientations and conformations of the antigens. It has been proposed that residues that have a direct role in antibody–antigen contact with one antibody may perform a steric role in stabilizing a specific conformation in recognition by another antibody (Obukhova et al. 2011).
The structure of the Fv portion of an anti-BGA antibody has been determined (Thomas et al. 2002) at 2.2 Å resolution, but complexes with the trisaccharide epitope fragment could not be obtained. The Fv was obtained from an scFv expressed from a synthetic gene based on the sequence of the anti-A monoclonal IgM, AC1001 (Chen et al. 1987). Its affinity for the antigen is low, Ka 3.4 × 103 M−1, but scanning of a glycan micro-array confirmed its specificity for the A trisaccharide. The antibody has a V-shaped pocket 11 Å deep (Figure 3M), 4.7 Å wide at the top and 2.8 Å wide at its base. This is well able to accommodate the key terminal GalNAc and docking experiments have shown how the GalNAc can specifically bind within it (Woods et al. 2013).
More recent docking and molecular simulations have produced a model for the recognition of the blood group A trisaccharide by BGA (Makeneni et al. 2014), which satisfactorily explained its specificity. A histidine residue is a notable feature of the site, and antigen binding was shown to be pH dependent. It increases at lower pH, which is the opposite behavior to that of the anti-Salmonella antibody Se155-4.
Antibodies to internal epitopes of bacterial heteropolysaccharides
This group of antibodies recognizes LPS antigens of Gram-negative bacterial pathogens, including epitopes from four O-chains and two cores. O-Chains are highly antigenic, and often contain immunodominant branching sugar repeating units. In the biosynthesis of O-chains, whether linear or branched, repeating units of two to five sugars are assembled as blocks and transferred to the growing chain (Raetz and Whitfield 2002). The epitope recognized by an antibody may encompass parts of more than one repeating unit, and it can be necessary to synthesize oligosaccharides that span through two units to establish which segment constitutes the epitope.
In addition to the antibodies considered here, two further examples should be noted. The original antigen for the phosphocholine-binding myeloma protein M603 is not known, but is probably of this class of antigen. The phosphocholine is bound in a centrally located pocket in the binding site, leaving a large surface surrounding it available for interaction with a carrier glycan (Satow et al. 1986). The structure of a human Fab to the capsular polysaccharide of Streptococcus pneumoniae type 23F has been deposited in the PDB as 4HIJ. It is the first example of an antibody structure for a Gram-positive antigen. The immunodominant moiety is a rhamnose branch, but further details await publication.
Salmonella typhimurium-specific antibody, Se155-4
The first high-resolution three-dimensional description of antibody recognition of any carbohydrate was monoclonal antibody Se155-4, an IgG1 λ1, in complex with a fragment of the O-chain from Salmonella typhimurium serogroup B (Cygler et al. 1991). The Fab from Se155-4 was successfully crystallized with a dodecasaccharide ligand obtained by phage degradation of the O-chain from S. typhimurium, which contains the repeating tetrasaccharide →2(Abeα1,3)Man-α1,4Rhaα1,3Galα1→ with a protruding abequose. The antibody combining site utilizes a combination of groove- and cavity-type recognition. Unambiguous electron density for the central trisaccharide segment showed, as expected, that the abequose was the dominant element (Figure 4A) with a contact surface contribution of 121 Å2 out of 255 Å2 in a central pocket ∼8 Å deep and 7 Å wide, formed chiefly by Trp residues and a Phe together with backbone segments. There are three His residues in the site, including one that forms a critical water bridge at the base of the abequose-binding pocket (Figure 4B), which explained the marked pH-dependence of binding (Bundle, Eichler, et al. 1994). CDRs H1, H3, L1 and L3 contribute to the pocket. The heptasaccharide complex suggested that the full antigen would be bound almost perpendicular to the Vh : Vl interdomain surface, though most groove-type sites follow this interface (Cygler et al. 1993). An intramolecular hydrogen bond seen in the trisaccharide antigen in the crystal structure could not be detected by NMR in the free trisaccharide, which is a strong indication that binding requires a change in conformation of the antigen by rotation around the glycosidic bonds (Bundle, Baumann, et al. 1994).
Antibody interactions with the epitope include numerous hydrophobic contacts and microcalorimetry has shown that there is a considerable positive entropic contribution to the binding of various ligands, which was attributed to the favorable desolvation of the hydrophobic protein surface (Sigurskjold and Bundle 1992). Being a dideoxy sugar, abequose is considerably more hydrophobic than most carbohydrates, and almost all of the side chains interacting with it are aromatic. This is reflected by the relative contributions of enthalpy and entropy to the overall binding, which vary greatly with temperature, and the Ka reaches a maximum of 1.6 × 105 M−1 at ∼290 K. These findings emphasize the complexity of associations between carbohydrate ligands and proteins generally, and how a full thermodynamic study can provide much greater insight than a simple measurement of Ka.
Se155-4 was also the first structurally characterized carbohydrate-specific antibody for which Fab and scFv fragments expressed from synthetic genes in E. coli were used to characterize the effects of amino acid residue mutations around the combining site (Brummell et al. 1993; Zdanov et al. 1994). For example, serial mutation of each of the four solvent exposed residues (Gly100, His101, Gly102 and Tyr103) on CDR H3 did not abolish binding. Substitutions involving the contact residue HisH101 with carboxylate or amide side chain groups resulted in binding affinities close to wild type, while substitutions of GlyH102 resulted in a significant decrease in affinity. There were no instances of rational site-directed mutagenesis in which the affinity for the antigen improved.
Bacteriophage display was used to produce a variety of altered antibody fragments (Deng et al. 1994). The mutants with improved binding were found to have CDR H2 side chains mutated to smaller ones, e.g., Ser to Ala, and Ala to Gly, suggesting that improvement in binding may have involved removal of steric clashes rather than enhanced positive interactions. A second round of phage display in which the mutations were limited to CDR residues generated an antibody with improved affinity that had a single Met to Ile replacement of a residue behind a binding-site Trp, which altered slightly the shape of the Abe pocket.
Shigella flexneri O-chain polysaccharide-specific antibody, SYA/J-6
The Sh. flexneri variant Y contains a linear tetrasaccharide repeat →2Rhaα1,2Rhaα1,3Rhaα1,3GlcNAcα1→ (Carlin et al. 1984; Hanna and Bundle 1993). A murine mAb, SYA/J-6 was generated in response to immunization with bacterial cells of Sh. flexneri variant Y (Carlin et al. 1984, 1986). The affinity of this antibody for a pentasaccharide in which the tetrasaccharide repeating unit is extended by a Rha at the reducing end was 3.8 × 104 M−1 (Vyas et al. 1993, 2002).
The structures of SYA/J-6 Fab were initially determined as complexes with the above pentasaccharide and Rha-2-deoxy-Rha-GlcNAc trisaccharide (Vyas et al. 1993). However, only two sugar rings were located at the time, the GlcNAc and its preceding Rha. The Rha is completely buried within the binding site and the GlcNAc is 75% buried. The second Rha appears to be mainly solvent exposed. Since then, the structures have been improved to 2.50 Å for the pentasaccharide and 2.30 Å for the trisaccharide, with unambiguous density for both antigens (Vyas et al. 2002). The antibody CDRs display high complementarity with the pentasaccharide (Figure 4C), making a total of eight hydrogen bonds and other contacts to all five carbohydrate residues (Figure 4D) along a deep groove (25 Å long, 10 Å deep at the center and 12 Å wide), with the largest number of polar contacts to the GlcNAc residue. Its Ka was 2.5 × 105 M−1 whereas the trisaccharide reached 1.7 × 106 M−1. The deep groove allowed the trisaccharide to achieve higher affinity where the central Rha residue is completely buried. In contrast, the ends of the tetrasaccharide repeat are solvent exposed. In a similar manner to the Sh. flexneri antibody F22-4, peptide mimics have also been produced for the Y-type antigen and solved as a complex with the antibody (Vyas et al. 2003). There were many water molecules mediating the peptide's interaction with the Fab, which may account for its affinity being only 2-fold higher than that of the pentasaccharide.
Shigella flexneri serotype 2a-specific antibody, F22-4
The Sh. flexneri serotype 2a is defined by the repeating branched pentasaccharide unit →2Rhaα1,2Rhaα1,3(Glcα1,4)Rhaα1,3GlcNAcβ1→ (Phalipon et al. 2006). A number of oligosaccharide fragments of this antigen have been synthesized (Belot et al. 2004; Wright et al. 2004; Phalipon et al. 2006). The development of an effective vaccine against this pathogen has proven difficult, with only a few glycoconjugates with these antigens observed to induce protective antibodies (Phalipon et al. 2006). Of five mAbs produced from immunization with homologous bacteria, an IgG1κ antibody, F22-4, displayed reactivity to various Shigella 2a O-antigen (O-Ag) constructs, including the pentasaccharide repeating unit and a decasaccharide. Inhibition assays with both the penta- and decasaccharide revealed high avidity binding (μm range) and subsequent isothermal calorimetry yielded a Ka of 1 × 106 M−1 (Theillet et al. 2009).
Crystal structures of F22-4 complexed with both penta- and decasaccharide units of the O-Ag have been recently determined to 1.8 and 2.0 Å, respectively (Vulliez-Le Normand et al. 2008). The binding site is a shallow groove (Figure 4E) 20 Å × 15 Å × 8 Å along the Vh/Vl boundary and a large area of 1125 Å2 is buried upon antigen binding. The structures show nine adjoining residues of the antigen forming a helical structure on the antibody surface (Figure 4F), with six residues making contact with antibody (Vulliez-Le Normand et al. 2008). The structure suggests that a minimal Shigella O-Ag epitope of two repeating units is required for optimal complementarity and an effective glycoconjugate vaccine. Like Se155-4, F22-4 uses a groove-type binding with a cavity to accommodate the branched residue with a complex pattern of 11 H-bonds and 14 coordinated water molecules (Figure 4G). Since the Shigella serotype Y and serotype 2a O-Ag have similar features, it is not surprising that antibodies raised against these structures share identical germ-line gene segments; however, the shorter CDR H3 of F22-4 (four amino acids) forms a deeper pocket for the branching Glc residue than SYA/J-6 antibody (nine amino acids). Additionally, a phage library was used to pan F22-4 against peptides that could bind and mimic the Shigella O-Ag. A decapeptide that could bind F22-4 with high affinity was identified and although the crystal structure revealed an entirely helical conformation of this peptide in the bound state, immunological mimicry was not achieved (Theillet et al. 2009). The solution NMR structure of the peptide alone revealed that the helical nature of the peptide was only maintained in the central region, which may account for the lack of mimicry.
Francisella tularensis-specific antibodies, Ab52 and N62
Francisella tularensis is a facultative intracellular pathogen requiring low infection dose upon inhalation with high mortality rates, making it a potential agent for bioterrorism (Conlan 2011). Its O-Ag consists of a tetrasaccharide repeat comprised of highly modified sugars such as galactosaminuronic acid amide, and is the dominant epitope of this pathogen; the repeating structure is –2Qui4NFoβ1,4GalNAcANα1,4GalNAcANα1,3QuiNAcβ1– (Gunn and Ernst 2007). Among a panel of antibodies, the IgG2a κ monoclonal Ab52 displayed highest avidity for this internal epitope spanning two tetrasaccharide O-Ag repeats. It conferred protection to BALB/c mice when injected with a lethal dose of live virulent F. tularensis (Roche et al. 2011; Lu et al. 2012).
The unliganded Fab structure of Ab52 was recently determined at 2.1 Å; experiments to obtain complexes with a dimer of the tetrasaccharide repeating unit were not successful (Rynkiewicz et al. 2012). The binding site has the form of a large canyon between the Vh and Vl domains with a central pocket (Figure 4H). The CDR H3 is unusually short, making it an ideal candidate for molecular docking simulations. Docking experiments performed included the evaluation of a range of O-Ag repeat lengths and arrangements, and the conformation of CDR H3 was predicted using the same method as for chP3 Fab above. Further, the simulation is subjected to grid-based ligand docking with energetics or glide algorithm, which systematically approximates conformational, orientational and positional space of the docked ligand, accompanied by a scoring function for the predicted affinity (Friesner et al. 2004). Though the minimal epitope was thought to be two repeats of the tetrasaccharide O-Ag, glide scores showed that the terminal A and D′ sugar residues were poorly accommodated in the binding site. Thus, an epitope consisting of a V-shaped hexasaccharide was proposed to bind Ab52 Fab.
The structure of an Fab of N62, an antibody specific for the terminal epitope of this antigen, has recently been solved by the same group (Lu et al. 2013) at 2.6 Å resolution without a ligand. N62 and Ab52 are the first pair of antibodies to be solved to internal and terminal epitopes of the same antigen. N62 has a cavity 8 × 10 × 9 Å (Figure 4I) lined with aromatic residues. Docking experiments suggested that it would accommodate a terminal Qui4NFo residue with the GalNAcAN contacting the rim of the cavity above it. The larger contact area provided by N62 burying a terminal residue compared with recognition of the internal epitope by Ab52 explains the higher affinity of N62 for its epitope, 1.9 × 105 M−1. Vaccines designed to present terminal epitopes rather than internal ones may therefore lead to antibodies that are more protective.
The inner core LPS-specific antibodies, WN1 222-5 and LPT3-1
LPS cores are attractive targets for antibody therapeutics and vaccines given their higher conservation relative to O-chains. WN1 222-5, an IgG2a κ, reacts with core region of LPS and has been shown to neutralize a broad range of pathogenic Gram-negative serovars even in the presence of the O-chain (Di Padova, Brade, et al. 1993; Di Padova, Gram, et al. 1993). Furthermore, WN1 222-5 has been shown to reduce the inflammatory cascade of septic shock in vivo, likely due to the hindering of LPS uptake and subsequent transfer to the Toll-like receptor 4-myloid differentiation factor 2 (TLR4-MD2) complex (Pollack et al. 1997).
The ligand used in the structural work on the Fab (Gomery et al. 2012) was a dodecasaccharide obtained from the core of E. coli R2 type, and is the largest bacterial oligosaccharide solved as an antibody complex (Figure 4J). Its structure is GlcNα1,2Glcα1,2-Glcα1,3(Galα1,6)Glcα1,3(Hepα1,7)(P4)Hepα1,3(P4)Hepα1,5(Kdoα2,4)Kdoα2,6(P4)GlcNβ1,6GlcNα1-P where Hep is l-glycero-d-manno-heptose and Kdo is 3-deoxy-d-manno-oct-2-ulosonic acid; the Ka for this antigen was 3.1 × 107 M−1, a very high value for a carbohydrate–protein interaction. The complex was solved at 1.73 Å resolution and the free antibody at 2.13 Å. The binding site is an open groove along the Vh–Vl interface with a protruding CDR H2 that moves upon binding the antigen, resulting in a total of 525 Å2 of buried surface area. Seven of the sugars formed contacts to the antibody (Figure 4K), including the five residues of the compact inner core region plus two from the adjacent outer core. The other three sugars of the outer core were also well resolved, but the di-GlcN phosphate of the lipid A portion was disordered. The majority of the binding interactions, 12 of 13 H-bonds, were with the Vh domain. The ligand also displayed 10 intramolecular bonds. The basis for TLR4 mimicry by WN1 222-5 is centered on the recognition of three key residues, Hep I, II and III, which form the majority of the epitope; these residues lie in the same conformation observed in the TLR4-LPS-MD2 structure (PDB code: 3FXI) (Park et al. 2009). Significantly, the attachment site for the O-chain on the GlcIII residue is not part of the epitope, but is solvent exposed and allows WN1 222-5 to overcome the heterogeneity associated with the O-chain. There are several phosphates around the core, but only the phosphate on the second heptose showed a significant charge interaction, with Arg52 of CDR H2.
A second anti-inner core mAb, LPT3-1 (IgG2b) has recently been determined to 2.69 Å resolution in complex with Neisseria meningitidis lipooligosaccharide (Parker et al. 2014). Neisseria meningitidis is typically comprised of an inner core tetrasaccharide GlcNAcα1,2Hepα1,3(P4)Hepα1,5Kdo, where the terminal α1,2GlcNAc is unique to N. meningitidis (Yang et al. 2012). Common substitutions on this conserved inner core include β1,4Glc on HepI, and ethanolamine phosphate substitution at 3-OH and/or 6-OH positions on HepII (Weynants et al. 2009; Yang et al. 2012). As with WN1 222-5, the lipid A carbohydrate moiety did not participate in binding and was only partially ordered in the complex structure of LPT3-1. The antibody displays groove type binding, with a total buried accessible surface area of 300 Å2 (Figure 4L), of which only 23 Å2 is provided by the light chain (Parker et al. 2014). TyrH50 and TrpL92 residues help in the formation of a hydrophobic pocket which accommodates the methyl group on the terminal GlcNAc residue. There are a total of nine hydrogen bonds, with the GlcNAcα1,2 Hep disaccharide forming most of these interactions (Figure 4M). The unliganded structure is not available for LPT3-1, and therefore conclusions cannot be made whether an induced fit mechanism is utilized during binding.
Antibodies to terminal epitopes on bacterial and fungal oligosaccharides
In addition to the anti-tumor antibodies, structures have been solved for antibodies that recognize non-reducing terminal epitopes on a variety of other glycan antigens, including bacterial polysaccharides, and fungal and human oligosaccharides. These epitopes often induce antibodies of higher affinity, perhaps because their presentation allows the antibody sites to develop larger contact surfaces, as noted for the F. tularensis antibodies above. However, they can also help to evade the immune response by mimicking host antigens, e.g., the blood group B epitope displayed by a Salmonella LPS (Perry and MacLean 1992).
The V. cholerae O1-specific antibody, S-20-4
Analysis of protective antibodies against the pathogenic V. cholerae serogroup O1 led to the conclusion that the Ogawa serotype displays a specific antigenic determinant (Apter et al. 1993), with ∼90% of the binding energy attributed to the terminal monosaccharide residue (Wang et al. 1998). The Inaba antigen consists of an α1,2-linked perosamine polysaccharide (perosamine being 4,6-dideoxy-4-amino-d-Man), N-acylated with 3-deoxy-l-glycero-tetronic acid, while the Ogawa antigen bears an additional O-methyl group at the C2 of the terminal sugar (Apter et al. 1993; Ito et al. 1994; Villeneuve et al. 2000). Only weak cross-reactivity to the Inaba antigen has been reported for Ogawa serotype-specific antibodies (Liao et al. 2002).
Crystal structures of an IgG1κ antibody, S-20-4, complexed with mono- and disaccharides of the Ogawa antigens were solved to 2.3 and 2.8 Å, respectively (Villeneuve et al. 2000). The structures revealed a central hydrophobic pocket defined by Trp and other aromatic residues that accommodate the methylated terminal perosamine residue with 420 Å2 of buried surface area (Figure 5A). The affinity of this antibody is relatively high, 3.9 × 105 M−1 for the monosaccharide and 1.2 × 106 M−1 for the disaccharide, which has been attributed to specific interactions with the methyl group. Intriguingly, the structure showed that the Inaba antigen would fit in the combining site of S-20-4 with identical contact sites as the Ogawa antigen (Figure 5B). Thus, the low cross-reactivity towards the Inaba antigen implies that dehydration of the hydrophobic pocket must be the driving force for binding to the Ogawa antigen.
Antibodies against chlamydial LPS antigens
Members of the bacterial family Chlamydiaceae possess an unusual truncated LPS consisting of lipid A linked to short oligomers of the inner core sugar αKdo that are joined by 2,4 and/or 2,8 linkages (Brade and Rietschel 1984; Brade et al. 1987). These carbohydrate structures were used to generate a large number of mAbs that fall largely into two distinct V-gene restricted families named after their prototypic clones, the S25-2 and S25-23 types. The S25-2 type displayed a striking range of specificities and cross-reactivities, while the S25-23 type bound exclusively to the Chlamydiaceae-specifc Kdo2,8Kdo2,4Kdo trisaccharide antigen (Brade et al. 1987, 1994, 2000; Kosma et al. 1990, 1999; Holst et al. 1991; Muller et al. 1997; Müller-Loennies et al. 2000).
Crystal structures of the S25-2 family of antibodies display a pocket of conserved V-gene sequence that binds terminal Kdo residues (Figure 5C) via several hydrogen bonds (Figure 5D) (Nguyen et al. 2003; Brooks et al. 2008). Significantly, a bifurcated salt bridge between an arginine of CDR H2 and the carboxyl group of the terminal Kdo was the first observation of a charged residue interaction in carbohydrate binding by an antibody (Nguyen et al. 2003). Additional Kdo residues of di- and trisaccharide chlamydia antigens as well as synthetic unnatural antigens are accommodated in the combining sites of S25-2 family mAbs by a flexible groove composed mainly of an Asn in CDR H2 and an Arg in L1 (Figure 5E), where the mutation in CDR H2 of AsnH53 to Lys allows additional interactions to Kdo oligosaccharides and increases avidity for all antigens (Blackler et al. 2011).
S25-2 family mAbs with different CDR H3s display varying levels of cross-reactivity to the 2,8- and 2,4-linked Kdo oligosaccharides. Whereas S25-2 and the similar S25-39 display higher avidity for the terminal 2,8 linkage, S45-18 and S54-10 (each using a different CDR H3 sequence) extend a phenylalanine residue from H3 into the binding pocket where it forms favorable stacking interactions against the terminal 2,4 linkage (Figure 5F) (Nguyen et al. 2003; Brooks, Müller-Loennies, et al. 2010). S73-2 also preferentially binds the 2,4 terminal linkage, with Kdo2,4Kdo2,4Kdo bound in the same orientation as S45-18 and S54-10, but shows weak cross-reactivity for the 2,8-terminal linkage, where Kdo2,8Kdo2,4Kdo adopts a “bent” conformation with Kdo2 bound in the conserved pocket (Figure 5G) (Brooks, Müller-Loennies, et al. 2010). S67-27 displays a long backward-leaning CDR H3 that allows the synthetic modified Kdo antigen Kdo2,8-7-O-Me-Kdo to bind with higher avidity than natural antigens through extra surface contact with CDR H3 (Figure 5H) (Brooks, Müller-Loennies, et al. 2010).
These structures suggest the S25-2 V-gene combination can provide recognition of a wide range of antigens of Kdo-like structure through a conserved cross-reactive binding site that is tuneable by CDR H3 to refine specificity, provide redundant specificity with unique binding mechanisms, or enable binding to previously unencountered antigen modifications. Immunization with Kdo antigens also produced antibodies of alternate light chain V-gene descent. S64-4 displays less cross-reactivity than the S25-2 family antibodies but much higher avidity for Kdo2,8Kdo2,4Kdo2,6GlcN4P1,6GlcN1P. The structure revealed a similar binding pocket for terminal Kdo with a more open binding site above that requires a large planar antigen conformation to achieve significant complementarity (Figure 5I; Evans et al. 2011). This mAb demonstrates the value of different V-gene combinations to recognize additional epitopes of the same antigen to enable tighter binding of larger ligands.
The S25-23-type antibodies use a different set of heavy chain V, D and J genes to confer strict recognition of Kdo2,8Kdo2,4Kdo containing antigens, with no observed cross-reactivity towards the Kdo mono- or disaccharide, or even the Kdo2,4Kdo2,4Kdo trisaccharide (Brade et al. 1997; Müller-Loennies et al. 2000). ITC of S25-23 showed high affinity with KD values of 6.54 × 10−8 and 9.90 × 10−8 for the Kdo2,8Kdo2,4Kdo tri- and Kdo2,8Kdo2,4Kdo2,6GlcN4P-1,6GlcN1P pentasaccharide, respectively (Haji-Ghassemi et al. 2014). Recently, the crystal structures of an S25-23-type antibody, S25-26 were reported to high resolution in liganded and unliganded forms (Haji-Ghassemi et al. 2014).
In contrast to the S25-2 family of mAbs, S25-26 does not have a specific pocket for the recognition of a single Kdo residue, but rather has an extended groove along the heavy chain that spreads the trisaccharide across the combining site to allow recognition of all three residues in a linkage/length-dependent manner (Figure 5J). This mode allows for hydrogen bonds (of which there are 13 total) to be evenly distributed among each of the three Kdo residues (Figure 5K), instead of the majority of binding occurring through the terminal Kdo residues as in the S25-2-type antibodies. Further, S25-26 relies more heavily on water to form a complementary surface to antigen with a total of 16 water bridge interactions from 10 water molecules (Figure 5K). The groove-type recognition of S25-26 allows for specific interactions to the first and terminal Kdo carboxyl groups, which in turn depends on the glycosidic linkage length and stereochemistry, explaining specificity toward the longer Kdo2,8Kdo2,4Kdo trisaccharide over the Kdo2,4Kdo2,4Kdo counterpart and the lack of observed cross-reactivity to Kdo mono and Kdo2,8Kdo disaccharide.
Finally, the structures of three unliganded forms as well as the liganded form of S25-26 revealed density for N-glycosylation present on Asn 85 of the variable heavy chain, adjacent to CDR H3. Analysis of the glycan revealed a heterogeneous mixture with a common root structure that contained an unusually high number of terminal Gal–Gal moieties, which have been implicated in allergic responses to therapeutic mAb treatment (Chung et al. 2008). One of the unliganded structures of S25-26 showed significant order of the glycan with appropriate electron density for nine residues. Interestingly, all S25-23 type (of which five have been characterized) antibodies possess the identical glycosylation site, and there is preliminary evidence that partial cleavage of the N-linked sugars results in increased affinity of Fab towards antigen (unpublished data).
The clear conclusion from this collection of structures is how the immune system has evolved to provide redundant and adaptable protection against important bacterial carbohydrates using a relatively small number of germ-line genes.
Mycobacterial LAM-specific antibody, CS-35
Because of their prominent role in pathogenesis of mycobacterial infections, surface glycolipids such as LAMs have been attractive targets for vaccination (Strohmeier and Fenton 1999). A large portion of LAMs contain α1,5-linked d-arabinofuranose (Araf) units, with branching points consisting of two arrangements: linear tetra-arabinoside, Araβ1,2Araα1,5Araα1,5Araα- and the related hexa-arabinoside Araβ1,2Araα1,5(Araβ1,2Araα1,3)Araα1,5Araα- (Nigou et al. 2003; Briken et al. 2004).
Several antibodies specific for LAMs have been identified from mycobacterial infections, of which mAb IgG CS-35 is the most extensively studied (Verbon et al. 1990; Kaur et al. 2002; Arias-Bouda et al. 2003). Binding assays with synthetic oligosaccharides indicated CS-35 has highest avidity for the hexasaccharide branched terminal structure (Ka 1 × 105 M−1) and to a lesser extent for the linear tetrasaccharide structure (Kaur et al. 2002). The antibody also loses avidity upon the addition of mannose residues capping the non-reducing terminal arabinosyl end, also known as ManLAMs.
Crystal structures of CS-35 Fab complexed with the tetra- and hexa-arabinosyl residues were solved to 1.8 and 2.0 Å resolution, respectively (Murase et al. 2009). The CDRs formed a triangular cavity with high complementarity to the non-reducing ends of the Y-shaped branched hexasaccharide epitope (Figure 5L), which wraps around TyrH98, and a groove complementary to the reducing end of the LAM structure (Figure 5M). All six CDRs contact the ligand and 509 Å2 of surface is buried. There are few direct H-bonds between the hexasaccharide ligand and the antibody, but there is an extensive network through water molecules, three of which are internal to the ligand. As with the other antibodies specific to terminal epitopes, CS-35 mAb displays the highest specificity to the terminal sugar residues (Figure 5M), evident from the clear electron density for these residues as well as the Ara at the branching point.
Candida albicans mannobiose-specific antibody, C3.1
Candida species are the most common cause of candidiasis, a fungal infection that occurs in immunocompromised patients, and the leading causative agent is C. albicans (MacCallum 2010). A few di- and trimannoside conjugates have elicited protective capabilities in animal models (Cutler 2005; Xin et al. 2008; Fidel and Cutler 2011). The best characterized mAb against the homopolymeric β-mannans of C. albicans is the protective mAb C3.1 (Han et al. 2000). Unlike earlier findings of Kabat with anti-dextran antibodies (Kabat 1960), C3.1 displayed reduced affinity as the size of the mannose oligomers increased past the trisaccharide unit.
STD NMR, chemical mapping, ELISA and computational methods using 1,2-β-linked mannose oligosaccharide fragments (Johnson et al. 2012) showed that oligosaccharides bound the antibody in conformations similar to the free ligand. The antigenic determinant is composed of β1,2Man disaccharide (Ka 3.7 × 104 M−1), with a marginal increase in affinity to the trisaccharide (Ka 5.5 × 104 M−1), and a decrease in affinity of the tetrasaccharide, (Ka 1.1 × 104 M−1). Based on the minimal epitope requirements of C3.1, a disaccharide protein-conjugate was designed and it conferred protective immunity against C. albicans in rabbits (Bundle et al. 2012; Johnson and Bundle 2013).
Modelling of C3.1 Fv using a structural homolog and the knowledge that it possesses a short (six residues) CDR H3, permitted the construction of a model that showed high complementarity to the natural C. albicans trisaccharide antigen. Further, the model revealed a groove type binding site, consistent with antibodies against internal homopolymeric epitopes.
Antibodies to homopolysaccharides
There are a number of areas where structural progress has been slow. This is particularly true where there are several identical overlapping antigenic determinants along the polysaccharide chain, and until recently, only unliganded structures had been obtained for antibodies to this type of antigen.
Early immunological studies of these antigens were facilitated by the relative ease with which defined oligosaccharide fragments can be obtained by acid hydrolysis. Experiments on antibodies specific for dextrans with Glc-oligosaccharides, in conjunction with similar experiments using oligopeptides, gave the first estimate for the size of the combining site of antibodies (Kabat 1957; Newman and Kabat 1985; Padlan and Kabat 1988). However, the apparent simplicity is deceptive, and the interpretation of binding data to long chain polysaccharides is not always straightforward. This is due to the mixture of complexes that can be formed with their oligosaccharides, with different occupancies of the subsites. For example, a trisaccharide may occupy subsites one through three, or subsites two through four. In addition to the numerous internal determinants along the chain, the non-reducing terminus can be a potent antigenic site, and while antibodies to the internal epitopes can also bind the terminal sugars, the reverse is generally not true.
Further, modelling of antibody–antigen complexes based on the structure of the antigen-free protein has been made particularly difficult by the fact that it cannot easily be determined in which direction the linear antigen will be bound in the site.
The galactan-specific myeloma protein, J539
J539 was one of the first anti-carbohydrate mouse myeloma proteins described and it is specific for galactans consisting of β1,6-linked galactose residues (Manjula et al. 1975). This antibody was induced by treatment of BALB/c mice with mineral oil (Jolley et al. 1973). Several other myeloma proteins (e.g., T601, X24, X44, S10 and T191) share similar galactan specificity and these proved to be formed from the same Vh and Vl genes with limited heterogeneity (Manjula et al. 1977; Rudikoff et al. 1980; Glaudemans 1987). The estimated Ka values for tri- and tetrasaccharide fragments were 1.5 and 3.4 × 105 M−1, respectively (Jolley et al. 1974). Measurements with other fragments and analogues led to a model in which there were four subsites within a surface binding cleft, with most of the binding energy deriving from interaction with the non-reducing terminal unit (Glaudemans et al. 1986; Glaudemans 1991).
The Fab fragment structure of the antibody was solved and refined to 4.5 Å by Navia et al. (1979), and improved successively to 2.6 Å resolution (Suh et al. 1986) and to 1.9 Å resolution (PDB entry: 2FBJ). The crystals were obtained only in the absence of oligosaccharide, and soaking experiments were unsuccessful due to the tight packing of the Fab fragments. The X-ray structure showed a large pocket (Figure 6A) connected to two grooves whose walls are defined by Trp and Tyr residues (Suh et al. 1986). Fitting of a pentasaccharide ligand on the basis of only van der Waals contacts was attempted, but it was complicated by the fact that the chain direction could not be established (a problem that was also encountered with modelling of the Brucella A system) and by the flexibility of the β1,6 linkage. While a consistent picture of the interactions of the sugars with the binding-site has been obtained, the physical locations of the postulated subsites remain unknown.
Brucella abortus O-chain polysaccharide-specific hybridoma, Yst9.1
Species of Brucella produce major cell wall polysaccharide antigens based on the unusual sugar N-formyl-perosamine (4,6 dideoxy-4-formamido-d-Man) (Caroff et al. 1984; Freer et al. 1995). In the B. abortus A polysaccharide, the perosamine residues are α1,2 linked with a small percentage of α1,3. In the B. mellitensis M antigen, every fifth linkage is an α1,3 (Bundle, Cherwonogrodzky, et al. 1987; Bundle, Cherwonogrodzky, Perry, et al. 1987; Peters et al. 1990). Recently, the M structure was re-investigated and the repeat unit was shown to have three α1,2 linkages to each α1,3 (Kubler-Kielb and Vinogradov 2013). The α1,2 linkage along with the d-manno configuration leads to the A antigen having a helical form with the α1,2 linkages forming a central axis, and the sugar rings perpendicular to it. The angle between successive formamido groups is 216°, giving rise to a five-residue conformational repeat along the helix (Kihlberg et al. 1991). Among panels of monoclonal antibodies raised against the two antigens, one termed Yst9.1, raised to the homogeneous form of the A antigen produced by Yersinia enterocolytica, had the greatest specificity for the A antigen (Bundle et al. 1984, 1989). Investigation with SPR revealed an avidity of 1.9 × 107 M−1 for the IgG, which is considerably higher than most carbohydrate-specific IgM immunoglobulins (Young et al. 1999).
A model of the antibody–antigen complex was proposed by Oomen et al. (1991). When compared with the subsequent crystal structure of the unliganded Fab, determined at 2.45 Å resolution (Rose et al. 1993), the model matched in many aspects except for a 3 Å translation of the V domains with respect to each other. The packing of the Fab molecules was such that their binding sites were occluded by neighboring molecules, so soaking experiments with oligosaccharides were unsuccessful. The binding site forms a large groove 20 Å long, 10 Å deep and 15 Å wide (Figure 6B), lined with aromatic residues. This crevice is broader than that of J539, as would be required for the greater width of the A antigen with its sugar rings perpendicular to the helix axis, compared with the extended form of the α1,6 galactans. The groove is clearly long enough to accommodate at least a pentasaccharide and the bulk of the interaction appears to be hydrophobic in nature. A central depression in the groove could accommodate a formamido group. A series of pentasaccharide analogs were synthesized in which N-formamido groups were selectively replaced with hydroxyl groups, i.e., N-formamido perosamine became d-Rha (Kihlberg and Bundle 1991; Kihlberg et al. 1991). However, binding studies of them could not be interpreted into a consistent set of subsites (D.R. Bundle, personal communication). A structure of an antigen complex is needed to elucidate the exact binding mechanism to this unique carbohydrate ligand.
The polysialic acid-specific antibody, mAb735
The bacteria N. meningitides and E. coli capsular type K1 are known agents of human bacterial meningitis. Both bacteria display capsular polysaccharide composed largely of α2,8-linked N-acetylneuraminic acid (also called group B meningitis polysaccharide or GBMP) (Jennings et al. 1985; Michon et al. 1985). Several antibodies specific for GBMP have been reported, most notable of which is mAb735 that recognizes homopolymers of GBMP (Frosch et al. 1985). As the binding sites of immunoglobulins had for years been established to be no longer than 8–10 saccharide residues, it was surprising to observe that mAb735 would only effectively bind GBMP polymers many times that length. Given that the antibody could not directly interact with potentially hundreds of required saccharide residues, the answer had to lie in the behavior of longer-chain polysaccharides.
NMR studies of the polysaccharide showed that it adopted extended helical conformations in solution with the carboxylate groups protruding on the “inside” of the helix and N-acetyl groups protruding on the “outside” (Michon et al. 1987; Wessels et al. 1987; Wessels and Kasper 1989; Brisson et al. 1992). The preponderance of the helix form increases with the number of oligosaccharide residues, which explains the preference of mAb735 for longer polymer lengths. Thermodynamics studies revealed two mAb735 Fabs binding per 41-residue polymer with an affinity of ∼3 × 105 M−1, which suggested an epitope of 10–20 residues, whereas oligosaccharides of 9–15 residues were bound an order of magnitude weaker by the IgG (Evans et al. 1995).
The Fab crystal structure of unliganded mAb735 was determined at 2.8 Å resolution (Evans et al. 1995) and revealed a combining site with complementarity in shape and charge to the predicted helical conformation of the antigen. Attempts to obtain crystals with PSA were not successful. A set of positively charged residues is clustered to one side of a shallow groove, while the other half is mainly hydrophobic. This would allow ionic interactions to occur with the negatively charged carboxyl groups at one end of the antigen helix, while the twist of the helix would bring the N-acetyl groups towards the hydrophobic residues.
Recently, a structure at a resolution of 1.8 Å was reported for an Fv form of mAb735 in a complex with an octasaccharide fragment of polysialic acid (Nagae et al. 2013). The hapten was bound in a unique manner, such that a U-shaped conformation at the interface between the binding sites of two Fv molecules was formed (Figure 6C), with two internal H-bonds bridging the arms and many associated water molecules. This conformation is unlikely to represent the situation in solution. Residues 2–4 and 6–8 contact the 2 Fvs, respectively, and 11 water molecules were involved in the H-bonding (Figure 6D). The area buried per trisaccharide element was 406 Å2 (Figure 6E) and the highest Ka measured by isothermal calorimetry was for a pentasaccharide, 1.5 × 103 M−1. CDR H3 formed most of the H-bonds, but all the CDRs except L3 contacted the antigen and there was no conformational change on binding. There were no charge interactions between amino acid side chains and the carboxylates of the octasaccharide.
Knowledge of the ability of high-molecular-weight GBMP to form extended helical structures in solution has formed the basis for new vaccines against human bacterial meningitis (Jennings 1997; Pon et al. 1997; Alfonso et al. 2002; Mond and Kokai-Kun 2008). Replacement of the acetyl groups by N-propyl ones led to an antigen with enhanced vaccine potential. An effect of GBMP chain length on immunogenicity was seen in immunization trials in mice using polysaccharide-tetanus toxoid conjugates with long NPrGBMP or short (NPrSia)4 glycans. The mAbs raised against the long NPrGBMP had greater bactericidal activity in vitro (Pon et al. 1997). Two antibodies to this antigen, 13D9 and 6B9, have been solved crystallographically at 2.45 and 2.06 Å, respectively (Johal et al. 2013). The structure of 13D9 revealed a CDR H2 that lies in a non-canonical conformation, which docking studies show is a critical feature in accommodating the extended NPrPSA antigen. A proposed model of extended NPr-PSA decasaccharide bound to 13D9 was consistent with STD-NMR experiments. Interestingly, chain direction was established due to the unfavorable binding energy calculated for one direction versus the other. Further, 13D9 displayed superior protective properties against meningitis and its convex surface was again consistent with binding of the antigen in an extended helical conformation. In contrast, the 6B9 site had a groove into which a shorter epitope would fit, but this size of epitope is not protective. Hence, the clinical effectiveness of GBMP-based vaccines will depend on their ability to induce antibodies that recognize extended helical GBMP epitopes on bacterial cell surfaces.
General features of anti-carbohydrate antibodies
The antigens for the antibodies covered here differ widely in size, from monosaccharides to dodecasaccharides, and in some cases include charged groups such as carboxylates and phosphates. Therefore, only broad conclusions about recognition factors can be drawn. Soon after the structures of Se155-4 and BR96 were reported, Wilson and Stanfield (1995) compared them with the classic predictions made by Kabat about groove and cavity sites. They found that his predictions were generally confirmed by the structures, but the sites were more complex. The H-bonds were formed by Asp, Asn, Glu, Gln and Arg residues and the peptide backbone, including bidentate bonds and coordinated water molecules. The aromatic residues Trp and Tyr lined the grooves and bound to the hydrophobic faces of the sugars. The variety of antigens reviewed here is a far wider one, but Wilson and Stanfield's conclusions remain valid. With regard to the types of amino acid that occur in binding sites, the same residues are common in the sites of other carbohydrate-binding proteins such as lectins; however, antibodies do not use calcium ions as C-type lectins and legume lectins do.
CDR H3 is the most important of the CDRs (Xu and Davis 2000), reflecting its unique genetic origin from the D gene segment. Its length and composition (Table II) helps to control the overall shape of the site and it contributes disproportionately to the binding of the antigen. There is a considerable spread in lengths of the CDR H3, from 2 to 13 residues, plus the exception of PG9 at 26 residues, with the commonest length being 6 or 7 residues. There is no apparent correlation between the size of the carbohydrate epitope and the length of the CDR H3, and while the majority of the antibodies show cleft-shaped-binding sites for the antigen, the CDR H3 length and sequence cause wide variations in this feature.
Antigen . | Antibody . | CDR H3 sequence . |
---|---|---|
F. tularensis | Ab52 | GF |
Sh. flexneri 2a | F22-4 | PM |
F. tularensis | N62 | YRF |
nPro-polysialic | 6B9 | LRAV |
Tn antigen | 237mAb | GKVRN |
Lex antigen | 291-2G3-A | ETGTRF |
V. cholera | S-20-4 | HFYAVL |
Polysialic acid | Mab735 | GGKFAM |
B. abortus | Yst9.1 | DPYGPA |
nPro-polysialic | 13D9 | SRGRTL |
Chlamydia | S25-26 | DETGSWF |
Mycobacteria | CS35 | FGNYVPF |
LPS core | WN1 222-5 | QGRGYTL |
Salmonella B | Se155-4 | GGHGYVG |
Sh. flexneri Y | SYA/J-6 | GGAVGAM |
Galactan | J539 | LHYYGYN |
Human ABO(H) blood group A | BGA | QYGNLWF |
N. meningitidis | LPT3-1 | MRITTDWF |
Ley antigen | BR96 | GLDDGAWF |
Gal ceramide | L363 | AAGIRWAWF |
Chlamydia | S25-2 | DHDGYYERF |
NeuGc-GM3 | P3 | SGVREGRAQWF |
GD3 glycolipid | R24 | GGTGTRSLYYF |
Chlamydia | S54-10 | DMRRFDDGDAM |
HIV-1 gp120 | 2G12 | KGSDRLSDNDPF |
Chlamydia | S73-2 | DINPGSDGYYDAL |
HIV-1 gp120 | PG9 | EAGGPDYRNGYNYYDF YDGYYNYHYM |
Antigen . | Antibody . | CDR H3 sequence . |
---|---|---|
F. tularensis | Ab52 | GF |
Sh. flexneri 2a | F22-4 | PM |
F. tularensis | N62 | YRF |
nPro-polysialic | 6B9 | LRAV |
Tn antigen | 237mAb | GKVRN |
Lex antigen | 291-2G3-A | ETGTRF |
V. cholera | S-20-4 | HFYAVL |
Polysialic acid | Mab735 | GGKFAM |
B. abortus | Yst9.1 | DPYGPA |
nPro-polysialic | 13D9 | SRGRTL |
Chlamydia | S25-26 | DETGSWF |
Mycobacteria | CS35 | FGNYVPF |
LPS core | WN1 222-5 | QGRGYTL |
Salmonella B | Se155-4 | GGHGYVG |
Sh. flexneri Y | SYA/J-6 | GGAVGAM |
Galactan | J539 | LHYYGYN |
Human ABO(H) blood group A | BGA | QYGNLWF |
N. meningitidis | LPT3-1 | MRITTDWF |
Ley antigen | BR96 | GLDDGAWF |
Gal ceramide | L363 | AAGIRWAWF |
Chlamydia | S25-2 | DHDGYYERF |
NeuGc-GM3 | P3 | SGVREGRAQWF |
GD3 glycolipid | R24 | GGTGTRSLYYF |
Chlamydia | S54-10 | DMRRFDDGDAM |
HIV-1 gp120 | 2G12 | KGSDRLSDNDPF |
Chlamydia | S73-2 | DINPGSDGYYDAL |
HIV-1 gp120 | PG9 | EAGGPDYRNGYNYYDF YDGYYNYHYM |
aThe sequences shown span from VH residue 95–100, numbered according to the PDB entries.
Antigen . | Antibody . | CDR H3 sequence . |
---|---|---|
F. tularensis | Ab52 | GF |
Sh. flexneri 2a | F22-4 | PM |
F. tularensis | N62 | YRF |
nPro-polysialic | 6B9 | LRAV |
Tn antigen | 237mAb | GKVRN |
Lex antigen | 291-2G3-A | ETGTRF |
V. cholera | S-20-4 | HFYAVL |
Polysialic acid | Mab735 | GGKFAM |
B. abortus | Yst9.1 | DPYGPA |
nPro-polysialic | 13D9 | SRGRTL |
Chlamydia | S25-26 | DETGSWF |
Mycobacteria | CS35 | FGNYVPF |
LPS core | WN1 222-5 | QGRGYTL |
Salmonella B | Se155-4 | GGHGYVG |
Sh. flexneri Y | SYA/J-6 | GGAVGAM |
Galactan | J539 | LHYYGYN |
Human ABO(H) blood group A | BGA | QYGNLWF |
N. meningitidis | LPT3-1 | MRITTDWF |
Ley antigen | BR96 | GLDDGAWF |
Gal ceramide | L363 | AAGIRWAWF |
Chlamydia | S25-2 | DHDGYYERF |
NeuGc-GM3 | P3 | SGVREGRAQWF |
GD3 glycolipid | R24 | GGTGTRSLYYF |
Chlamydia | S54-10 | DMRRFDDGDAM |
HIV-1 gp120 | 2G12 | KGSDRLSDNDPF |
Chlamydia | S73-2 | DINPGSDGYYDAL |
HIV-1 gp120 | PG9 | EAGGPDYRNGYNYYDF YDGYYNYHYM |
Antigen . | Antibody . | CDR H3 sequence . |
---|---|---|
F. tularensis | Ab52 | GF |
Sh. flexneri 2a | F22-4 | PM |
F. tularensis | N62 | YRF |
nPro-polysialic | 6B9 | LRAV |
Tn antigen | 237mAb | GKVRN |
Lex antigen | 291-2G3-A | ETGTRF |
V. cholera | S-20-4 | HFYAVL |
Polysialic acid | Mab735 | GGKFAM |
B. abortus | Yst9.1 | DPYGPA |
nPro-polysialic | 13D9 | SRGRTL |
Chlamydia | S25-26 | DETGSWF |
Mycobacteria | CS35 | FGNYVPF |
LPS core | WN1 222-5 | QGRGYTL |
Salmonella B | Se155-4 | GGHGYVG |
Sh. flexneri Y | SYA/J-6 | GGAVGAM |
Galactan | J539 | LHYYGYN |
Human ABO(H) blood group A | BGA | QYGNLWF |
N. meningitidis | LPT3-1 | MRITTDWF |
Ley antigen | BR96 | GLDDGAWF |
Gal ceramide | L363 | AAGIRWAWF |
Chlamydia | S25-2 | DHDGYYERF |
NeuGc-GM3 | P3 | SGVREGRAQWF |
GD3 glycolipid | R24 | GGTGTRSLYYF |
Chlamydia | S54-10 | DMRRFDDGDAM |
HIV-1 gp120 | 2G12 | KGSDRLSDNDPF |
Chlamydia | S73-2 | DINPGSDGYYDAL |
HIV-1 gp120 | PG9 | EAGGPDYRNGYNYYDF YDGYYNYHYM |
aThe sequences shown span from VH residue 95–100, numbered according to the PDB entries.
It is interesting to compare the CDR H3 compositions to those of all six CDRs combined from a large set of mouse antibodies, and to their framework regions (Table III; Padlan 1990). Compared with the full CDR set, the anti-carbohydrate CDR H3 set (213 residues) possesses far more Gly (17.4% vs. 6.8%) but far less Ser (3.3% vs. 14.7%). Aromatic residues form 23.5% of the H3 set compared with 18.7% of the CDR set and only 9.8% of the framework set; in the H3 CDRs, there is more Phe (8.5 vs. 3.4%) but similar amounts of Trp and Tyr. Asp and Glu both occur more often than their amide forms and are unequal in abundance in the H3 set (∼4 : 1). Though these four residues were found to be important for H-bonding to antigens (Wilson and Stanfield 1995), they only comprise 17.3% of the H3 residues. Notably Asn is only 4.2% compared with 8.3% in the full CDR set. Charged residues comprise 23% of the total and the Arg content of H3 is higher than the CDR and framework sets (8.0% vs. 3.8% and 3.4%, respectively). Lys is very low in H3 (1.4% vs 4.4% and 5.44%), Thr is less common (4.2% vs 6.6% and 9.37%) and Cys is completely absent. As would be expected for external loops, the hydrophobic amino acids Leu, Val, Ile, Met and Pro together are only 14% of all CDR H3 residues. Hence, judging by their H3 compositions only, carbohydrate antigens appear to induce antibodies with CDR compositions that are different from those for other antigens. But it is the higher Phe, Arg, Asp and Gly contents that are distinctive, and the H-bonding residues Asn, Ser and Thr usually associated with carbohydrate binding are actually less common. It has been shown that the D segments are most often incorporated into Vh genes in reading frames which favor Gly and Tyr residues (Abergel and Claverie 1991). This is consistent with the observed Gly content but not as much with the Tyr.
Amino acid . | CDR H3 . | CDR H3 (%) . | CDR set (%) . | FR set (%) . |
---|---|---|---|---|
Ala | 14 | 6.5 | 5.29 | 6.65 |
Arg | 17 | 8.0 | 3.76 | 3.36 |
Asn | 9 | 4.2 | 8.32 | 1.06 |
Asp | 20 | 9.4 | 5.78 | 3.48 |
Cys | 0 | 0 | 0.05 | 2.40 |
Gln | 3 | 1.4 | 3.43 | 5.94 |
Glu | 5 | 2.3 | 2.41 | 4.33 |
Gly | 37 | 17.4 | 6.75 | 10.44 |
His | 5 | 2.3 | 2.83 | 0.34 |
Ile | 3 | 1.4 | 3.03 | 4.23 |
Leu | 10 | 4.7 | 4.45 | 8.72 |
Lys | 3 | 1.4 | 4.40 | 5.44 |
Met | 7 | 3.3 | 2.27 | 1.39 |
Phe | 18 | 8.5 | 3.37 | 3.32 |
Pro | 7 | 3.3 | 3.48 | 3.94 |
Ser | 7 | 3.3 | 14.70 | 12.12 |
Thr | 9 | 4.2 | 6.61 | 9.37 |
Trp | 7 | 3.3 | 2.18 | 2.52 |
Tyr | 25 | 11.7 | 13.13 | 3.97 |
Val | 7 | 3.3 | 3.78 | 6.98 |
Total | 213 | (6298) | (17,142) |
Amino acid . | CDR H3 . | CDR H3 (%) . | CDR set (%) . | FR set (%) . |
---|---|---|---|---|
Ala | 14 | 6.5 | 5.29 | 6.65 |
Arg | 17 | 8.0 | 3.76 | 3.36 |
Asn | 9 | 4.2 | 8.32 | 1.06 |
Asp | 20 | 9.4 | 5.78 | 3.48 |
Cys | 0 | 0 | 0.05 | 2.40 |
Gln | 3 | 1.4 | 3.43 | 5.94 |
Glu | 5 | 2.3 | 2.41 | 4.33 |
Gly | 37 | 17.4 | 6.75 | 10.44 |
His | 5 | 2.3 | 2.83 | 0.34 |
Ile | 3 | 1.4 | 3.03 | 4.23 |
Leu | 10 | 4.7 | 4.45 | 8.72 |
Lys | 3 | 1.4 | 4.40 | 5.44 |
Met | 7 | 3.3 | 2.27 | 1.39 |
Phe | 18 | 8.5 | 3.37 | 3.32 |
Pro | 7 | 3.3 | 3.48 | 3.94 |
Ser | 7 | 3.3 | 14.70 | 12.12 |
Thr | 9 | 4.2 | 6.61 | 9.37 |
Trp | 7 | 3.3 | 2.18 | 2.52 |
Tyr | 25 | 11.7 | 13.13 | 3.97 |
Val | 7 | 3.3 | 3.78 | 6.98 |
Total | 213 | (6298) | (17,142) |
Amino acid . | CDR H3 . | CDR H3 (%) . | CDR set (%) . | FR set (%) . |
---|---|---|---|---|
Ala | 14 | 6.5 | 5.29 | 6.65 |
Arg | 17 | 8.0 | 3.76 | 3.36 |
Asn | 9 | 4.2 | 8.32 | 1.06 |
Asp | 20 | 9.4 | 5.78 | 3.48 |
Cys | 0 | 0 | 0.05 | 2.40 |
Gln | 3 | 1.4 | 3.43 | 5.94 |
Glu | 5 | 2.3 | 2.41 | 4.33 |
Gly | 37 | 17.4 | 6.75 | 10.44 |
His | 5 | 2.3 | 2.83 | 0.34 |
Ile | 3 | 1.4 | 3.03 | 4.23 |
Leu | 10 | 4.7 | 4.45 | 8.72 |
Lys | 3 | 1.4 | 4.40 | 5.44 |
Met | 7 | 3.3 | 2.27 | 1.39 |
Phe | 18 | 8.5 | 3.37 | 3.32 |
Pro | 7 | 3.3 | 3.48 | 3.94 |
Ser | 7 | 3.3 | 14.70 | 12.12 |
Thr | 9 | 4.2 | 6.61 | 9.37 |
Trp | 7 | 3.3 | 2.18 | 2.52 |
Tyr | 25 | 11.7 | 13.13 | 3.97 |
Val | 7 | 3.3 | 3.78 | 6.98 |
Total | 213 | (6298) | (17,142) |
Amino acid . | CDR H3 . | CDR H3 (%) . | CDR set (%) . | FR set (%) . |
---|---|---|---|---|
Ala | 14 | 6.5 | 5.29 | 6.65 |
Arg | 17 | 8.0 | 3.76 | 3.36 |
Asn | 9 | 4.2 | 8.32 | 1.06 |
Asp | 20 | 9.4 | 5.78 | 3.48 |
Cys | 0 | 0 | 0.05 | 2.40 |
Gln | 3 | 1.4 | 3.43 | 5.94 |
Glu | 5 | 2.3 | 2.41 | 4.33 |
Gly | 37 | 17.4 | 6.75 | 10.44 |
His | 5 | 2.3 | 2.83 | 0.34 |
Ile | 3 | 1.4 | 3.03 | 4.23 |
Leu | 10 | 4.7 | 4.45 | 8.72 |
Lys | 3 | 1.4 | 4.40 | 5.44 |
Met | 7 | 3.3 | 2.27 | 1.39 |
Phe | 18 | 8.5 | 3.37 | 3.32 |
Pro | 7 | 3.3 | 3.48 | 3.94 |
Ser | 7 | 3.3 | 14.70 | 12.12 |
Thr | 9 | 4.2 | 6.61 | 9.37 |
Trp | 7 | 3.3 | 2.18 | 2.52 |
Tyr | 25 | 11.7 | 13.13 | 3.97 |
Val | 7 | 3.3 | 3.78 | 6.98 |
Total | 213 | (6298) | (17,142) |
However, when the H-bond forming residues in the structures solved with bound carbohydrate are tallied up, a total of 94 residues, a different picture emerges. Overall, the VH domain predominates with almost twice as many H-bonding residues than the Vl, 61 vs. 33, divided equally among its CDRs; L3 is a little higher at 24. This distribution differs from the earlier analysis of Padlan (1994) in which the six CDRs were more equal in their contributions. The most abundant H-bonding residue is Tyr (14 occurrences), followed by Asn (12) and Trp (11). Also common are Arg (10), His (9) and Glu (8). Further, the distribution of these residues among the six CDRs is strikingly uneven. Arg and Asn chiefly occur in the CDRs L3 and H2, 4–5 times each; Tyr and Trp in L3 (9) and H1 (11); His in L1 and H3, four times in each; and Glu in H2 (6). Given that Tyr and Trp are most important in forming the binding-site cavities and consequently making van der Waals contacts with an antigen, their abundance is to be expected.
Another aspect of recognition that deserves attention is the buried surface area of antigen binding, which varies considerably for each antibody. The buried surface areas of the various HIV-1-specific antibodies have been reviewed elsewhere (Kong et al. 2014). As a further example, the areas of the chlamydial LPS-binding antibodies discussed in this review are summarized in Table IV. Generally, the S25-2 type antibodies in complex with ligands larger than a disaccharide occlude ∼300 Å2, with near even contributions from light and heavy chains. However, S64-4 forms a complementary surface that encloses 382 Å2 of the pentasaccharide ligand, owing to a different light chain that forms additional contacts to the lipid A carbohydrate backbone. In contrast, the heavy chain-dependent S25-26 antibody forms a shallow groove (when compared with the S25-2 type), yielding a binding surface area of 331 Å2. The higher affinity of S67-27 for the synthetic disaccharide Kdo2,8-7-O-Me-Kdo compared with Kdo2,8Kdo is consistent with the 21 Å2 increase in buried surface area for the 7-O-Me.
Antibody (PDB code) . | Total for antigen: from light/heavy (Å2) . | Antigen . |
---|---|---|
S25-2 (3T4Y) | 147: 82.0/64.8 | Kdo |
S25-2 (3SY0) | 286: 154/132 | Kdo2,8Kdo2,4Kdo |
S54-10 (3I02) | 315: 144/171 | Kdo2,4Kdo2,4Kdo |
S73-2 (3HZV) | 311: 174/137 | Kdo2,8Kdo2,4Kdo |
S67-27 (3IKC) | 295: 145/150 | Kdo2,8-7-O-Me-Kdo |
S67-27 (3IJY) | 274: 132/142 | Kdo2,8Kdo |
S64-4 (3PHO) | 382: 207/175 | Kdo2,8Kdo2,4Kdo2, 6GlcN4P1,6GlcN1P |
S25-26 (4M7J) | 331: 47.2/284 | Kdo2,8Kdo2,4Kdo |
Antibody (PDB code) . | Total for antigen: from light/heavy (Å2) . | Antigen . |
---|---|---|
S25-2 (3T4Y) | 147: 82.0/64.8 | Kdo |
S25-2 (3SY0) | 286: 154/132 | Kdo2,8Kdo2,4Kdo |
S54-10 (3I02) | 315: 144/171 | Kdo2,4Kdo2,4Kdo |
S73-2 (3HZV) | 311: 174/137 | Kdo2,8Kdo2,4Kdo |
S67-27 (3IKC) | 295: 145/150 | Kdo2,8-7-O-Me-Kdo |
S67-27 (3IJY) | 274: 132/142 | Kdo2,8Kdo |
S64-4 (3PHO) | 382: 207/175 | Kdo2,8Kdo2,4Kdo2, 6GlcN4P1,6GlcN1P |
S25-26 (4M7J) | 331: 47.2/284 | Kdo2,8Kdo2,4Kdo |
Calculated with AreaIMol (Lee and Richards 1971) in CCP4 suite (Bailey 1994) using 1.4 Å probe radius and standard van der Waals radii. All solvent molecules were excluded for the calculations.
Antibody (PDB code) . | Total for antigen: from light/heavy (Å2) . | Antigen . |
---|---|---|
S25-2 (3T4Y) | 147: 82.0/64.8 | Kdo |
S25-2 (3SY0) | 286: 154/132 | Kdo2,8Kdo2,4Kdo |
S54-10 (3I02) | 315: 144/171 | Kdo2,4Kdo2,4Kdo |
S73-2 (3HZV) | 311: 174/137 | Kdo2,8Kdo2,4Kdo |
S67-27 (3IKC) | 295: 145/150 | Kdo2,8-7-O-Me-Kdo |
S67-27 (3IJY) | 274: 132/142 | Kdo2,8Kdo |
S64-4 (3PHO) | 382: 207/175 | Kdo2,8Kdo2,4Kdo2, 6GlcN4P1,6GlcN1P |
S25-26 (4M7J) | 331: 47.2/284 | Kdo2,8Kdo2,4Kdo |
Antibody (PDB code) . | Total for antigen: from light/heavy (Å2) . | Antigen . |
---|---|---|
S25-2 (3T4Y) | 147: 82.0/64.8 | Kdo |
S25-2 (3SY0) | 286: 154/132 | Kdo2,8Kdo2,4Kdo |
S54-10 (3I02) | 315: 144/171 | Kdo2,4Kdo2,4Kdo |
S73-2 (3HZV) | 311: 174/137 | Kdo2,8Kdo2,4Kdo |
S67-27 (3IKC) | 295: 145/150 | Kdo2,8-7-O-Me-Kdo |
S67-27 (3IJY) | 274: 132/142 | Kdo2,8Kdo |
S64-4 (3PHO) | 382: 207/175 | Kdo2,8Kdo2,4Kdo2, 6GlcN4P1,6GlcN1P |
S25-26 (4M7J) | 331: 47.2/284 | Kdo2,8Kdo2,4Kdo |
Calculated with AreaIMol (Lee and Richards 1971) in CCP4 suite (Bailey 1994) using 1.4 Å probe radius and standard van der Waals radii. All solvent molecules were excluded for the calculations.
Water molecules play multiple roles in carbohydrate binding and they are of critical importance in the overall energetics. Though the number of water molecules identified depends on the resolution of the structure, many of the water networks around the reviewed ligands are extensive, particularly for the larger oligosaccharides. Notable examples include the mAb 735 Fv structure and the Sh. flexneri antibody F22-4. The network can adjust to aid the binding of related oligosaccharides, as seen in the case of the mycobacterial antibody CS-35, and this flexibility may come into play when a whole polysaccharide antigen is being bound rather than a fragment of it. The set of antibodies includes examples of critical water molecules forming part of H-bonding networks deep within the site, e.g., Se155-4, as well as waters H-bonding to the periphery around it. Where unliganded structures are available, it is apparent that the hydroxyls of the glycan ligand often displace bound water molecules. But simultaneously with these waters being returned to the bulk solvent, other water molecules are being captured that were not observed in the unliganded structures. Hence, assessing the overall entropy changes associated with waters during binding is far from simple. Another factor in binding energetics is the presence of internal H-bonds in the bound forms of some antigens, some involving water molecules such as the mycobacterial antibody CS-35. These help to form the antigens into compact shapes, but NMR experiments suggest that they may not be present in the free solution state.
A further energetic factor is induced fitting of the antibody site to the antigen. The prevalence of this factor is hard to assess because free and bound forms are not available for most of the antibodies, and intermolecular contacts in the crystals can also cause movements of the CDRs. But for some antibodies there are definite major shifts of the CDRs consistent with induced fitting to the antigen, including the anti-LeY antibody BR96 and the anti-Chlamydia antibodies. Similar changes were seen with antibodies to peptides and to DNA (Wilson and Stanfield 1993). In contrast, there were no changes seen in the mAb 735 Fv structure with octasialic acid compared with the unliganded Fab one.
In summary it is clear that antibodies against carbohydrates show distinct features; however, there are no “rules” governing their behavior and for each generalization that is made, there will be one or more antibodies that show contradictory behavior. For example, the much longer epitope requirement by the mAb735 and the pH sensitivity of Se155-4 are contrary to the behavior of other systems. Collectively, however, they give a generally consistent view of how recognition operates in these systems at the oligosaccharide level and there are indications that the recognition of the complete polysaccharide has additional features.
Funding
This work was supported by the Natural Sciences and Engineering Research Council of Canada.
Conflict of interest statement
None declared.
Abbreviations
Abe, 3,6-dideoxy-d-xylo-hexopyranose; CDRs, complementary determining loops; Fv, fragment variable; Ley, Lewis Y antigen; scFvs, single-chain fragment variables; sdAb, single domain antibody; SPR, surface plasmon resonance; STD, saturation transfer difference; Vh, variable domain of the heavy chain; Vl, variable domain of the light chain.
References
Author notes
The atomic coordinates and structure factors for all structures discussed are available in the Protein Data Bank, Research Collaborators for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org) (see Table I).